Archives
-
Tariffs and trade: How might partnerships change as tensions persist?
On McKinsey Perspectives
Brought to you by Alex Panas, global leader of industries, & Axel Karlsson, global leader of functional practices and growth platforms
Welcome to the latest edition of Only McKinsey Perspectives. We hope you find our insights useful. Let us know what you think at Alex_Panas@McKinsey.com and Axel_Karlsson@McKinsey.com.
—Alex and Axel
—Edited by Belinda Yu, editor, Atlanta
This email contains information about McKinsey's research, insights, services, or events. By opening our emails or clicking on links, you agree to our use of cookies and web tracking technology. For more information on how we use and protect your information, please review our privacy policy.
You received this email because you subscribed to the Only McKinsey Perspectives newsletter, formerly known as Only McKinsey.
Copyright © 2025 | McKinsey & Company, 3 World Trade Center, 175 Greenwich Street, New York, NY 10007
by "Only McKinsey Perspectives" <publishing@email.mckinsey.com> - 11:08 - 24 Jul 2025 -
Explore excellent PCB and PCBA solutions to help your project succeed!
Dear Info,
I am pleased to introduce our company, a trusted manufacturer of PCB, PCBA and PCB assembly services. With years of experience, we specialize in providing customized, high-quality products to meet the diverse needs of industries such as automotive, consumer electronics and telecommunications.
Our services cover everything from design and prototyping to full production. With state-of-the-art facilities and a focus on reliability, we ensure fast turnaround times and competitive pricing to effectively support your project.
I would love the opportunity to discuss how we can help you with your next project. Please let me know a convenient time to contact you, or feel free to contact us for more details.
Best regards,
WELLDONE ELECTRONICS
PCB + PCBA Manufacturer
Mobile: +86 138 0883 6312
Email:steven@welldonepcb.com
by "Tafuabi Maho" <mahotafuabi@gmail.com> - 06:36 - 24 Jul 2025 -
Struggling with AI adoption?
Struggling with AI adoption?
Struggling with AI adoption?
Hi Md Abul,
AI adoption brings technical, ethical, and cost challenges, making it tough to stay in control. We get it.
Enterprises crave centralized governance of AI tools to ensure long-term sustainability.
But how can they get there?
Get an insight into the challenges of shadow and agentic AI and how to overcome them in our free guide.
AI, APIs, and AI readiness:
The strategic blueprint for enterprise
Make this a strategic blueprint for your enterprise, and be on your way to AI readiness.
No forms or downloading - just instant access to the PDF
We hope you find this useful!
Yours,
Budha
Tyk, Huckletree 199 Bishopsgate, Broadgate, London, City of London EC2M 3TY, United Kingdom, +44 (0)20 3409 1911
by "Budha from Tyk" <communities@tyk.io> - 11:50 - 24 Jul 2025 -
Consistency and Partition Tolerance: Understanding CAP vs PACELC
Consistency and Partition Tolerance: Understanding CAP vs PACELC
Forwarded this email? Subscribe here for moreLatest articles
If you’re not a subscriber, here’s what you missed this month.
A Guide to Database Replication: Key Concepts and Strategies
A Guide to Database Transactions: From ACID to Concurrency Control
Database Schema Design Simplified: Normalization vs Denormalization
To receive all the full articles and support ByteByteGo, consider subscribing:
Modern databases don’t run on a single box anymore. They span regions, replicate data across nodes, and serve millions of queries in parallel.
However, every time a database tries to be fast, available, and correct at once, something has to give. As systems scale, the promise of fault tolerance collides with the need for correctness. For example, a checkout service can’t afford to double-charge a user just because a node dropped off the network. But halting the system every time a replica lags can break the illusion of availability. Latency, replica lag, and network partitions are not edge cases.
Distributed databases have to manage these trade-offs constantly. For example,
A write request might succeed in one region but not another.
A read might return stale data unless explicitly told to wait.
Some systems optimize for uptime and accept inconsistencies. Others block until replicas agree, sacrificing speed to maintain correctness.
Two models help make sense of this: the CAP theorem and the PACELC theorem. CAP explains why databases must choose between staying available and staying consistent in the presence of network partitions. PACELC extends that reasoning to the normal case: even without failure, databases still trade latency for consistency.
In this article, we will look at these two models as they apply to real-world database design and understand the various trade-offs involved.
Consistency and Partition Tolerance...
Continue reading this post for free in the Substack app
Like
Comment
Restack
© 2025 ByteByteGo
548 Market Street PMB 72296, San Francisco, CA 94104
Unsubscribe
by "ByteByteGo" <bytebytego@substack.com> - 11:36 - 24 Jul 2025 -
Boost Recovery and Wellness with Red Light Therapy
Dear Info,
We’re excited to introduce our Red Light Therapy Panels, designed to accelerate wound healing, relieve joint and muscle pain, and reduce fatigue. These panels also enhance energy levels, strengthen immunity, and improve sleep quality.
Whether for fitness centers, wellness clinics, or personal use, our panels provide a holistic approach to health, powered by advanced Quad-Wave technology for maximum benefits.
Let us know if you'd like to learn more or discuss potential collaborations. I’m happy to assist!
Best regards,
Sherry Zhong
Omedic Technolog (Dongguan)Co.,ltd
Whatsapp : +86 13172450763
by "Marilynn Mcconico" <mcconicomarilynn@gmail.com> - 11:17 - 24 Jul 2025 -
Cool off in the summer: Explore the charm of rechargeable misting fans
Dear Info,
We are a factory located in Jiangmen, Guangdong, and would like to introduce our Rechargeable Mist Fan KN-1169.
This fan comes with a rechargeable battery and offers adjustable wind speeds. It includes a built-in mist function, an LED display, adjustable height and swing, a night light, and a timer. It supports USB charging and has a large water tank, making it suitable for both indoor and outdoor use.
If you are interested or need more details, please feel free to contact us.
Best regards,
Joye
Kennede Electronics MFG Co., Ltd.
(Professional manufacturer of small home appliance)
※※※※※※※※※※※※※※※※※※※※
Mobile / Wechat/ Whatsapp : 86 13427437046
E-mail: export41@kennede.com
website: www.kennede.com
Address: No.21 Jintong Road, Tangxia Town,Pengjiang District, Jiangmen City, Guangdong Province , China.
Post Code: 529085
by "Younis Baz" <younisbaz590@gmail.com> - 08:31 - 24 Jul 2025 -
Digital upskilling: How companies can future-proof their workforce
On McKinsey Perspectives
Brought to you by Alex Panas, global leader of industries, & Axel Karlsson, global leader of functional practices and growth platforms
Welcome to the latest edition of Only McKinsey Perspectives. We hope you find our insights useful. Let us know what you think at Alex_Panas@McKinsey.com and Axel_Karlsson@McKinsey.com.
—Alex and Axel
•
The value of digital skills. The rise of AI and other technologies promises to boost business growth and productivity. But to achieve those benefits, companies must ensure that workers across the organization, in and out of tech, are digitally fluent, Senior Partner Brooke Weddle and coauthors say. McKinsey research finds that companies with the strongest digital and AI capabilities significantly outperform their less-advanced peers, with two to six times higher total shareholder returns. For organizations to perform at their best, all employees—including leaders—need to be techies.
—Edited by Belinda Yu, editor, Atlanta
This email contains information about McKinsey's research, insights, services, or events. By opening our emails or clicking on links, you agree to our use of cookies and web tracking technology. For more information on how we use and protect your information, please review our privacy policy.
You received this email because you subscribed to the Only McKinsey Perspectives newsletter, formerly known as Only McKinsey.
Copyright © 2025 | McKinsey & Company, 3 World Trade Center, 175 Greenwich Street, New York, NY 10007
by "Only McKinsey Perspectives" <publishing@email.mckinsey.com> - 11:04 - 23 Jul 2025 -
Reshaping the consumer sector with tech
Intersection
Share this email
Technology—especially automation—is affecting nearly every industry, and the consumer sector is no exception, say McKinsey’s Alexis Krivkovich, David Fuller, and Shaun Callaghan and their coauthors. To learn more about how companies can seize the opportunity at hand and reshape the consumer landscape, check out the latest edition of the Five Fifty.
Share these insights
Did you enjoy this newsletter? Forward it to colleagues and friends so they can subscribe too. Was this issue forwarded to you? Sign up for it and sample our 40+ other free email subscriptions here.
This email contains information about McKinsey's research, insights, services, or events. By opening our emails or clicking on links, you agree to our use of cookies and web tracking technology. For more information on how we use and protect your information, please review our privacy policy.
You received this email because you subscribed to the Intersection newsletter.
Copyright © 2025 | McKinsey & Company, 3 World Trade Center, 175 Greenwich Street, New York, NY 10007
by "McKinsey Quarterly Five Fifty" <publishing@email.mckinsey.com> - 04:13 - 23 Jul 2025 -
Thank you for supporting ByteByteGo Newsletter
Thank you for reading ByteByteGo Newsletter. As a token of our appreciation, we're offering you a limited-time offer of 20% off a paid subscription.
Redeem special offerHere are the benefits you unlock with a paid subscription:
- An extra deep dive on Thursdays
- Full archive
- Many expense it with team's learning budget
Join the thousands of other readers who pay for full access to ByteByteGo Newsletter - redeem your special offer today!
Thanks again for reading.
© 2025 ByteByteGo
548 Market Street PMB 72296, San Francisco, CA 94104
by "ByteByteGo" <bytebytego@substack.com> - 01:26 - 23 Jul 2025 -
The Growing AI Trust Gap in Cybersecurity
The Growing AI Trust Gap in Cybersecurity
Hi MD Abul,
There was a story this week that stopped me mid-scroll.
According to a new report, while 83% of cybersecurity executives believe AI is essential to staying competitive, only 10% of analysts trust AI to work independently and just 22% say it actually improves their productivity.
That’s a serious disconnect.
And it’s one I’ve seen show up far beyond cybersecurity.Here’s what’s happening:
Leadership sees AI as a breakthrough. Automation, efficiency, proactive threat detection it all sounds like progress.
But frontline teams?
They’re the ones dealing with false positives, shallow context, and black-box models that make decisions without clear reasoning.And that tension between vision and reality is one of the most overlooked challenges in AI adoption.
This isn't just a tech problem.
It’s an alignment problem.If the people closest to the work don’t trust the tools, they’ll either avoid them or overcorrect them. In either case, you lose the value.
So what can you do?
-
Start with explainability. Don’t just deploy AI help your teams understand what it’s doing and why.
-
Keep humans in the loop. Especially in areas like cybersecurity, AI should augment, not replace.
-
Use real success stories to build confidence. Trust grows when people see how it helps them today, not someday.
AI adoption isn’t just about capability. It’s about credibility.
And that starts on the ground floor.If you’re interested in exploring more of these challenges in depth, we’ll be covering them live this August at our Digital Stratosphere event.
Use the code "earlybird" for 20% off your ticket(s)!
Bonus: We’re booking private advisory sessions during the event, just 15 minutes of real talk with someone who’s done this before. Reply if you’d like us to connect you with the right person.
Best regards,
Eric Kimberling
by "Eric Kimberling" <eric.kimberling@thirdstage-consulting.com> - 08:31 - 23 Jul 2025 -
-
Premium Alternators & Starters for Commercial Vehicles from Superb Auto Electrical
Dear Info,
As a specialized manufacturer of auto electrical components since 1997, Superb Auto Electrical supplies OEM-grade alternators, starters, and sub-components to global distributors. Our Wuxi-based factory holds certifications for Bosch, Denso, Hitachi, and Nikko-compatible systems, with dedicated lines for commercial vehicles (trucks, forklifts) and specialty applications (marine, agriculture).
Key technical highlights include:
Advanced Solutions: 24V/48V/72V alternators (e.g., Isuzu NPR, Mitsubishi buses), Denso hairpin-wound stators
Quality Assurance: NSK bearings, Transpo regulators, 220°C-rated copper windings
Industry Coverage: US/European aftermarket support with ECE/SAE compliance
Attached catalog details 800+ SKUs for trucks and industrial equipment. For volume inquiries, kindly share your requirements.
Best regards,
Jocelyn
Superb Auto Electrical Co.,Ltd
Add:CD, 3rd Floor, No. 1 Commercial Building, 248 Dong Lan Road, Shanghai, China.
Mobile: +86-15000544435( Whatsapp)
Skype:info_888262
www.tech-superb.com
https://superbauto.en.alibaba.com
by "Galante Odette" <galanteodette@gmail.com> - 05:28 - 23 Jul 2025 -
Building a deep bench of future leaders
On McKinsey Perspectives
Brought to you by Alex Panas, global leader of industries, & Axel Karlsson, global leader of functional practices and growth platforms
Welcome to the latest edition of Only McKinsey Perspectives. We hope you find our insights useful. Let us know what you think at Alex_Panas@McKinsey.com and Axel_Karlsson@McKinsey.com.
—Alex and Axel
•
Critical leadership skills. In today’s unpredictable business climate, leading is more difficult now than ever before. CEOs must address a range of compounding challenges, and the highest-performing leaders also understand that these obstacles can bring opportunities. Global Managing Partner Bob Sternfels, Senior Partner Daniel Pacthod, and Senior Adviser David H. Berger explain that in these highly uncertain times, there are six key traits—such as continuous learning and selfless leadership—that enable CEOs to lead calmly, confidently, and with an eye toward the future.
—Edited by Belinda Yu, editor, Atlanta
This email contains information about McKinsey's research, insights, services, or events. By opening our emails or clicking on links, you agree to our use of cookies and web tracking technology. For more information on how we use and protect your information, please review our privacy policy.
You received this email because you subscribed to the Only McKinsey Perspectives newsletter, formerly known as Only McKinsey.
Copyright © 2025 | McKinsey & Company, 3 World Trade Center, 175 Greenwich Street, New York, NY 10007
by "Only McKinsey Perspectives" <publishing@email.mckinsey.com> - 01:34 - 23 Jul 2025 -
Try Pro Suite free for 30 days and get 30% off when you buy
Try Pro Suite Free and Save 30%
Start your trial As your business grows, you need tools that grow with it. Take your business to the next level with an easy-to-use, comprehensive CRM suite that has all the marketing, sales, and service tools you need to scale as your needs grow.
With Salesforce Pro Suite, you can:
- Grow faster with greater customisation and automation
- Equip sales teams with quoting and forecasting
-
Engage customers in real time across support channels
No credit card required.
- Offer valid until 31 July 2025
© 2025, Salesforce, Inc.
Salesforce.com 2 Silom Edge, 14th Floor, Unit S14001-S14007, Silom Road, Suriyawong, Bangrak, Bangkok 10500
General Enquiries: +66 2 430 4323
This email was sent to info@learn.odoo.com
Manage Preferences or Unsubscribe | Privacy Statement
Powered by Salesforce Marketing Cloud
by "Salesforce" <apacemarketing@mail.salesforce.com> - 12:07 - 23 Jul 2025 -
How Nubank Uses AI Models to Analyze Transaction Data for 100M Users
How Nubank Uses AI Models to Analyze Transaction Data for 100M Users
Forwarded this email? Subscribe here for moreFree NoSQL Training – and a Book by Discord Engineer Bo Ingram (Sponsored)
Get practical experience with the strategies used at Discord, Disney, Zillow, Tripadvisor & other gamechangers
July 30, 2025
Whether you’re just getting started with NoSQL or looking to optimize your NoSQL performance, this event is a fast way to learn more and get your questions answered by experts.
You can choose from two tracks:
Essentials: NoSQL vs SQL architectures, data modeling fundamentals, and building a sample high-performance application with ScyllaDB.
Advanced: Deep dives into application development practices, advanced data modeling, optimizing your database topology, monitoring for performance, and more.
This is live instructor-led training, so bring your toughest questions. You can interact with speakers and connect with fellow attendees throughout the event.
Bonus: Registrants get the ScyllaDB in Action ebook by Discord Staff Engineer Bo Ingram.
Disclaimer: The details in this post have been derived from the articles shared online by the Nubank Engineering Team. All credit for the technical details goes to the Nubank Engineering Team. The links to the original articles and sources are present in the references section at the end of the post. We’ve attempted to analyze the details and provide our input about them. If you find any inaccuracies or omissions, please leave a comment, and we will do our best to fix them.
Understanding customer behavior at scale is one of the core challenges facing modern financial institutions. With millions of users generating billions of transactions, the ability to interpret and act on this data is critical for offering relevant products, detecting fraud, assessing risk, and improving user experience.
Historically, the financial industry has relied on traditional machine learning techniques built around tabular data. In these systems, raw transaction data is manually transformed into structured features (such as income levels, spending categories, or transaction counts) that serve as inputs to predictive models.
While this approach has been effective, it suffers from two major limitations:
Manual feature engineering is time-intensive and brittle, often requiring domain expertise and extensive trial-and-error.
Generalization is limited. Features designed for one task (for example, credit scoring) may not be useful for others (for example, recommendation or fraud detection), leading to duplicated effort across teams.
To address these constraints, Nubank adopted a foundation model-based approach, which has already been transforming domains like natural language processing and computer vision. Instead of relying on hand-crafted features, foundation models are trained directly on raw transaction data using self-supervised learning. This allows them to automatically learn general-purpose embeddings that represent user behavior in a compact and expressive form.
The objective is ambitious: to process trillions of transactions and extract universal user representations that can power a wide range of downstream tasks such as credit modeling, personalization, anomaly detection, and more. By doing so, Nubank aims to unify its modeling efforts, reduce repetitive feature work, and improve predictive performance across the board.
In this article, we look at Nubank’s system design for building and deploying these foundation models. We will trace the complete lifecycle from data representation and model architecture to pretraining, fine-tuning, and integration with traditional tabular systems.
Overall Architecture of Nubank’s System
Nubank’s foundation model system is designed to handle massive volumes of financial data and extract general-purpose user representations from it. These representations, also known as embeddings, are later used across many business applications such as credit scoring, product recommendation, and fraud detection.
The architecture is built around a transformer-based foundation model and is structured into several key stages, each with a specific purpose.
1 - Transaction Ingestion
The system starts by collecting raw transaction data for each customer. This includes information like the transaction amount, timestamp, and merchant description. The volume is enormous, covering trillions of transactions across more than 100 million users.
Each user has a time-ordered sequence of transactions, which is essential for modeling spending behavior over time. In addition to transactions, other user interaction data, such as app events, can also be included.
2 - Embedding Interface
Before transactions can be fed into a transformer model, they need to be converted into a format the model can understand. This is done through a specialized encoding strategy. Rather than converting entire transactions into text, Nubank uses a hybrid method that treats each transaction as a structured sequence of tokens.
Each transaction is broken into smaller elements:
Amount Sign (positive or negative) is represented as a categorical token.
Amount Bucket is a quantized version of the transaction amount (to reduce numeric variance).
Date tokens such as month, weekday, and day of month are also included.
Merchant Description is tokenized using standard text tokenizers like Byte Pair Encoding.
See the diagram below:
This tokenized sequence preserves both the structure and semantics of the original data, and it keeps the input length compact, which is important because attention computation in transformers scales with the square of the input length.
3 - Transformer Backbone
Once transactions are tokenized, they are passed into a transformer model. Nubank uses several transformer variants for experimentation and performance optimization.
The models are trained using self-supervised learning, which means they do not require labeled data. Instead, they are trained to solve tasks like:
Masked Language Modeling (MLM), where parts of the transaction sequence are hidden and the model must predict them.
Next Token Prediction (NTP), where the model learns to predict the next transaction in the sequence.
More on these in later sections. The output of the transformer is a fixed-length user embedding, usually taken from the final token's hidden state.
4 - Self-Supervised Training
The model is trained on large-scale, unlabeled transaction data using self-supervised learning objectives. Since no manual labeling is required, the system can leverage the full transaction history for each user. The models learn useful patterns about financial behavior, such as spending cycles, recurring payments, and anomalies, by simply trying to predict missing or future parts of a user’s transaction sequence. As a simplified example, the model might see “Coffee, then Lunch, then…”. It tries to guess “Dinner”.
The size of the training data and model parameters plays a key role. As the model scales in size and the context window increases, performance improves significantly. By constantly guessing and correcting itself across billions of transactions, the model starts to notice patterns in how people spend money.
For instance, switching from a basic MLM model to a large causal transformer with optimized attention layers resulted in a performance gain of over 7 percentage points in downstream tasks.
5 - Downstream Fine-Tuning and Fusion
After the foundation model is pre-trained, it can be fine-tuned for specific tasks. This involves adding a prediction head on top of the transformer and training it with labeled data. For example, a credit default prediction task would use known labels (whether a customer defaulted) to fine-tune the model.
To integrate with existing systems, the user embedding is combined with manually engineered tabular features. This fusion is done in two ways:
Late Fusion uses models like LightGBM to combine embeddings and tabular data, but the two are trained separately.
Joint Fusion trains the transformer and the tabular model together in an end-to-end fashion using a deep neural network, specifically the DCNv2 architecture.
More on these in a later section.
6 - Centralized Model Repository
To make this architecture usable across the company, Nubank has built a centralized AI platform.
This platform stores pretrained foundation models and provides standardized pipelines for fine-tuning them. Internal teams can access these models, combine them with their features, and deploy fine-tuned versions for their specific use cases without needing to retrain everything from scratch.
This centralization accelerates development, reduces redundancy, and ensures that all teams benefit from improvements made to the core models.
Transforming Transactions into Model-Ready Sequences
To train foundation models on transaction data, it is critical to convert each transaction into a format that transformer models can process.
There are two main challenges when trying to represent transactions for use in transformer models:
Mixed data types: A single transaction includes structured fields (like amount and date) and textual fields (like the merchant name). This makes it harder to represent consistently using either a pure text or a pure structured approach.
High cardinality and cold-start issues: Transactions can be extremely diverse. Even simple changes like different merchant names, locations, or amounts result in new combinations. If each unique transaction is treated as an individual token, the number of tokens becomes enormous. This leads to two problems:
The embedding table grows too large to train efficiently.
The model cannot generalize to new or rare transactions it has not seen before, which is known as the cold-start problem.
To address these challenges, multiple strategies were explored for turning a transaction into a sequence of tokens that a transformer can understand.
Approach 1: ID-Based Representation (from Recommender Systems)
In this method, each unique transaction is assigned a numerical ID, similar to techniques used in sequential recommendation models. These IDs are then converted into embeddings using a lookup table.
While this approach is simple and efficient in terms of token length, it has two major weaknesses:
The total number of unique transaction combinations is extremely large, making the ID space impractical to manage.
If a transaction was not seen during training, the model cannot process it effectively. This makes it poorly suited for generalization and fails under cold-start conditions.
Approach 2: Text-Is-All-You-Need
This method treats each transaction as a piece of natural language text. The transaction fields are converted into strings by joining the attribute name and its value, such as "description=NETFLIX amount=32.40 date=2023-05-12", and then tokenized using a standard NLP tokenizer.
This representation can handle arbitrary transaction formats and unseen data, which makes it highly generalizable.
However, it comes at a high computational cost.
Transformers process data using self-attention, and the computational cost of attention increases with the square of the input length. Turning structured fields into long text sequences causes unnecessary token inflation, making training slower and less scalable.
Approach 3: Hybrid Encoding Scheme (Chosen by Nubank)
To balance generalization and efficiency, Nubank developed a hybrid encoding strategy that preserves the structure of each transaction without converting everything into full text.
Each transaction is tokenized into a compact set of discrete fields:
Amount Sign Token: One token represents whether the transaction is positive (like a deposit) or negative (like a purchase).
Amount Bucket Token: The absolute value of the amount is placed into a quantized bucket, and each bucket is assigned its token. This reduces the range of numeric values into manageable categories.
Date Tokens: Separate tokens are used for the month, day of the month, and weekday of the transaction.
Description Tokens: The merchant description is tokenized using a standard subword tokenizer like Byte Pair Encoding (BPE), which breaks the text into common fragments. This allows the model to recognize patterns in merchant names or transaction types.
By combining these elements, a transaction is transformed into a short and meaningful sequence of tokens. See the diagram below:
This hybrid approach retains the key structured information in a compact format while allowing generalization to new inputs. It also avoids long text sequences, which keeps attention computation efficient.
Once each transaction is tokenized in this way, the full transaction history of a user can be concatenated into a sequence and used as input to the transformer. Separator tokens are inserted between transactions to preserve boundaries, and the sequence is truncated at a fixed context length to stay within computational limits.
Training the Foundation Models
Once transactions are converted into sequences of tokens, the next step is to train a transformer model that can learn patterns from this data.
As mentioned, the Nubank engineering team uses self-supervised learning to achieve this, which means the model learns directly from the transaction sequences without needing any manual labels. This approach allows the system to take full advantage of the enormous volume of historical transaction data available across millions of users.
Two main objectives are used:
Next Token Prediction (NTP): The model is trained to predict the next token in a transaction sequence based on the tokens that came before it. This teaches the model to understand the flow and structure of transaction behavior over time, similar to how language models predict the next word in a sentence.
Masked Language Modeling (MLM): In this method, some tokens in the sequence are randomly hidden or “masked,” and the model is trained to guess the missing tokens. This forces the model to understand the surrounding context and learn meaningful relationships between tokens, such as the connection between the day of the week and spending type or between merchant names and transaction amounts.
See the diagram below:
Blending Sequential Embeddings with Tabular Data
While foundation models trained on transaction sequences can capture complex behavioral patterns, many financial systems still rely on structured tabular data for critical features.
For example, information from credit bureaus, user profiles, or application forms often comes in tabular format. To make full use of both sources (sequential embeddings from the transformer and existing tabular features), it is important to combine them in a way that maximizes predictive performance.
This process of combining different data modalities is known as fusion.
Nubank explored two main fusion strategies: late fusion, which is easier to implement but limited in effectiveness, and joint fusion, which is more powerful and trains all components together in a unified system.
See the diagram below:
Late Fusion (Baseline Approach)
In the late fusion setup, tabular features are combined with the frozen embeddings produced by a pretrained foundation model. Think of it like taking the “frozen feeling” from the pre-trained model (embeddings) and combining it with the checklist of facts such as age, credit score, profile details, and so on.
These combined inputs are then passed into a traditional machine learning model, such as LightGBM or XGBoost. While this method is simple and leverages well-established tools, it has an important limitation.
Since the foundation model is frozen and trained separately, the embeddings do not adapt to the specific downstream task or interact meaningfully with the tabular data during training. As a result, there is no synergy between the two input sources, and the overall model cannot fully optimize performance.
Joint Fusion (Proposed Method)
To overcome this limitation, Nubank developed a joint fusion architecture.
This approach trains the transformer and the tabular model together in a single end-to-end system. By doing this, the model can learn to extract information from transaction sequences that complements the structured tabular data, and both components are optimized for the same prediction task.
To implement this, Nubank selected DCNv2 (Deep and Cross Network v2) as the architecture for processing tabular features. DCNv2 is a deep neural network specifically designed to handle structured inputs. It combines deep layers with cross layers that capture interactions between features efficiently.
See the diagram below:
Conclusion
Nubank’s initiative to use foundational models represents a significant leap forward in how financial institutions can understand and serve their customers. By moving away from manually engineered features and embracing self-supervised learning on raw transaction data, Nubank has built a modeling system that is both scalable and expressive.
A key part of this success lies in how the system has been integrated into Nubank’s broader AI infrastructure. Rather than building isolated models for each use case, Nubank has developed a centralized AI platform where teams can access pretrained foundation models. These models are trained on massive volumes of user transaction data and are stored in a shared model repository.
Teams across the company can choose between two types of models based on their needs:
Embedding-only models, which use the user representation generated from transaction sequences.
Blended models, which combine these embeddings with structured tabular features using the joint fusion architecture.
This flexibility is critical.
Some teams may already have strong tabular models in place and can plug in the user embeddings with minimal changes. Others may prefer to rely entirely on the transformer-based sequence model, especially for new tasks where historical tabular features are not yet defined.
The architecture is also forward-compatible with new data sources. While the current models are primarily trained on transactions, the design allows for the inclusion of other user interaction data, such as app usage patterns, customer support chats, or browsing behavior.
In short, Nubank’s system is not just a technical proof-of-concept. It is a production-ready solution that delivers measurable gains across core financial prediction tasks.
References:
Understanding Our Customer’s Finances Through Foundation Models
Defining an Interface between Transaction Data and Foundational Models
Shipping late? DevStats shows you why. (Sponsored)
Still pretending your delivery issues are a mystery? They’re not. You’re just not looking in the right place.
DevStats gives engineering leaders brutal clarity on where delivery breaks down, so you can fix the process instead of pointing fingers.
✅ Track DORA and flow metrics like a grown-up
✅ Spot stuck work, burnout risks, and aging issues
✅ Cut cycle time without cutting corners
✅ Ship faster. With fewer surprises.
More AI tools won’t fix your delivery. More Clarity will.
SPONSOR US
Get your product in front of more than 1,000,000 tech professionals.
Our newsletter puts your products and services directly in front of an audience that matters - hundreds of thousands of engineering leaders and senior engineers - who have influence over significant tech decisions and big purchases.
Space Fills Up Fast - Reserve Today
Ad spots typically sell out about 4 weeks in advance. To ensure your ad reaches this influential audience, reserve your space now by emailing sponsorship@bytebytego.com.
© 2025 ByteByteGo
548 Market Street PMB 72296, San Francisco, CA 94104
Unsubscribe
by "ByteByteGo" <bytebytego@substack.com> - 11:34 - 22 Jul 2025 -
Tips to Revitalize Sales Transformation
Hi Info,
Struggling with transformer performance or customization?
Many businesses face challenges like inefficiency, high costs, and lack of tailored solutions. These issues can hurt productivity and growth.
At Wenzhou Kunji Trading Co., Ltd., we design and produce high-quality, customized transformers to meet your exact needs. With in-house R&D, technical expertise, and self-operated exports, we ensure reliable, cost-effective solutions delivered on time.Let’s discuss how we can help—reply now, and we’ll get started!
Best Regards,
Amber
Sales and Marketing Department
Email:amber@kunji-electric.com
info@kunji-electric.com
Website:www.kjtransformer.com
Wenzhou Kunji Trading Co., Ltd.
by "Laiq Hag" <pekikiebach406@gmail.com> - 07:58 - 22 Jul 2025 -
Make every sitting a pleasure: The secret of Runzhi office chair
Dear Manager,
Nice to Meet you in 2025!
This is Maymay from RZ furniture Co.,Ltd in China Foshan. (One stop manufacturer- - a manufacturing supplier).
Glad to know you are in market of Office Furniture. We are now for Mesh chair, Meeting chair, Ergonomic chair, Training chair, Room chairs, PU leather chairs, and furniture parts. We specialize in this field for Twelve years.
Contact me if any question and wonders. I will give you prompt reply.
Thanks for your patience.
Yours Sincerely
Bella: +8613924825558 bella@rzfurniture.com
Leah: +8613929999480 tan.leah9480@gmail.com
Maymay: +8618934386507 Lina01@rzfurniture.com
Ruizhi Furniture (Foshan) Co.,Ltd.
Web: https://www.rzfurniture.com/Office-Chair-pl6912256-p2.html 【Home】https://rzfurniture.en.alibaba.com/?spm=a2700.29482153.0.0.73f471d29GfxOM 【Shop】
Ins:maymay08152024
Add: Xinchong Industrial Park, Shangdong Tengjiao Area,Jiujiang Town,Nanhai District,Foshan ,Guangdong China.
by "Cilli Augusto" <cilliaugusto556@gmail.com> - 03:27 - 22 Jul 2025 -
Making sense of the trends shaping today’s labor market
On McKinsey Perspectives
Brought to you by Alex Panas, global leader of industries, & Axel Karlsson, global leader of functional practices and growth platforms
Welcome to the latest edition of Only McKinsey Perspectives. We hope you find our insights useful. Let us know what you think at Alex_Panas@McKinsey.com and Axel_Karlsson@McKinsey.com.
—Alex and Axel
•
Keeping options open. With so many labor market trends colliding, from AI and aging workforces to changing economic policy, job seekers and employers alike face high levels of uncertainty. In an interview with McKinsey talent leaders Brooke Weddle and Bryan Hancock, Indeed’s Chief Economist Svenja Gudell shares that in this unpredictable environment, employers are looking to be flexible by creating more part-time or contract roles rather than committing to full-time hires.
•
AI’s effect on workers. Based on an analysis of nearly 3,000 skills, “we were able to see that gen AI does not replace any job completely,” Gudell explains on an episode of McKinsey Talks Talent. Instead, gen AI is augmenting jobs, particularly for knowledge workers such as accountants and marketers. In the case of agentic AI, though, she notes that there are early signs of displacement: fewer listings for internships and entry-level jobs, for example. Listen to the full episode to discover the labor market trends that business leaders should be thinking about.
—Edited by Belinda Yu, editor, Atlanta
This email contains information about McKinsey's research, insights, services, or events. By opening our emails or clicking on links, you agree to our use of cookies and web tracking technology. For more information on how we use and protect your information, please review our privacy policy.
You received this email because you subscribed to the Only McKinsey Perspectives newsletter, formerly known as Only McKinsey.
Copyright © 2025 | McKinsey & Company, 3 World Trade Center, 175 Greenwich Street, New York, NY 10007
by "Only McKinsey Perspectives" <publishing@email.mckinsey.com> - 01:16 - 22 Jul 2025 -
How Strategy Champions win
Share this email
Brought to you by Andy West (Andy_West@McKinsey.com) and Dago Diedrich (Dago_Diedrich@McKinsey.com), global leaders of McKinsey’s Strategy & Corporate Finance Practice
New from McKinsey Quarterly
How Strategy Champions win
Unlock 12 building blocks 
This email contains information about McKinsey’s research, insights, services, or events. By opening our emails or clicking on links, you agree to our use of cookies and web tracking technology. For more information on how we use and protect your information, please review our privacy policy.
You received this email because you subscribed to our McKinsey Quarterly alert list.
Copyright © 2025 | McKinsey & Company, 3 World Trade Center, 175 Greenwich Street, New York, NY 10007
by "McKinsey & Company" <publishing@email.mckinsey.com> - 12:44 - 21 Jul 2025
