Siddharth Kumar from Intuit talks about increasing personal finance with AI

In the fourth edition of the Machine Learning Developer Summit (MLDS), Siddharth Kumar of Intuit discussed in depth the features, metrics, challenges and potential of developing a personal financial management app like Intuit Mint (formerly, targeting users in the US and Canada.

As Intuit’s Principal Data Scientist, Kumar is responsible for scaling Intuit’s personal finance platform, Mint. At Intuit, he has helped launch several successful AI-powered offerings, including cash flow forecasting, automated subscription discovery, automated budgeting to maximize savings, transaction categorization, anomaly spending detection, and automated customer groups.

What is mint?

With 4 million active users, Mint helps its users manage spending, budget, subscriptions, etc. As a personal financial management app, Mint holds more than a decade of user transaction data for tens of millions of users. Transaction data includes bank accounts, money management accounts, retirement or investment accounts, credit cards, trading and other financial services. Besides these, other data also includes clickstream, demographic, geographic location, and derived features such as aggregations and sequences.

Making decisions and predicting outcomes with this data is easier said than done. Kumar explained the strong and growing competition in the space, along with challenges such as data completeness, the dynamic nature of transaction data, user fatigue and legacy systems.

“We had to deal with a lot of legacy systems, which we moved to AWS,” Kumar said. Furthermore, they used EC2 inference nodes to score in real time (tens of millions of transactions logged per day), along with EMR or groups of nodes for better processing (training up to 100 seconds and millions of ML models in a few hours or Several x 1000 core sets), on-device or Unified Learning (iOS), etc. To store data, Intuit Mint currently uses Hive, DynamoDB, Redshift, and S3. In addition, the company makes use of Python, Pyspark, R, SQL, and SWIFT (OS) in terms of programming languages.

(Source: Intuit / MLDS)


Primary business metrics include Voice of the Customer (VoC), retention, click-through rate (CTR), net promoter score (NPS), profit and loss (PnL), relative ratings, and other capabilities in relation to competition.

AI metrics include: accuracy (cash flow); completeness (discovery of recurring transactions); Accuracy, recall and AUC such as detection of credit default and anomaly; cross-entropy (classification); Provision (MCMC) to improve savings; AIC, KL-divergence (especially aggregation) etc.

Potential AI accessories are everywhere

“Fintech is the next big thing. So, we all know that. The next thing is to automate customer ranking and relative spending order,” Kumar said.

Ability (Along with the technologies usedUse cases include credit default/late payment forecasting (logistical), cash flow forecasting (SVM), large transaction anomaly detection/alert (GB), transaction grading (NN), engagement maximization via path recommendation (boost), trade switching (GMM) + MAB), Spend Recommendations (Boost), Trend Patterns and Detailed Analytics for User or Overall (Statistical), Improved Financial Wellness (MAB), CLTV (Survival), and Optimize Marketing and Advertising Targeting (NN).

Kumar explained the technologies behind automated customer groups and relative spending ordering (auto-encryption + GMM), billing or subscription identifiers (FFTs), can I afford it? (RF), Automated Budgeting (combination + custom optimizer).

“If we give our customers a simple way to ask Mint if they can afford a particular purchase at a particular category or merchant, Mint will be able to advise based on users’ current spending pattern, which will help them expedite spending decisions.” Mint uses on-device RF (Swift, CoreML, iOS) for feedback learning and a multivariate statistical model for seeding.

Leave a Comment