project

GateKeep

Agentic personal finance

Built with
JavaScript · Agents
Source
Public on GitHub
Author
Samir Sengupta

An agentic-AI finance tracker that ingests transactions and reasons over them with tool-using agents to surface spend insights.

The problem

Budgeting apps categorise transactions. They do not reason about them. Knowing you spent $412 on “Shopping” is data; knowing that it repeats every month and is drifting upward is an insight, and that gap is where the useful part lives.

How it works

  • Transactions are ingested and handed to tool-using agents that can query, aggregate, and compare rather than simply label.
  • Agents reason over spending history to surface patterns — recurrence, drift, and outliers — instead of producing another pie chart.
  • The agent layer is the product: the interesting engineering is in giving a model the right tools and the right guardrails over personal financial data.

Why it exists

A working demonstration that agentic AI is most useful where the question is open-ended and the data is boring.

More work

Hiring for AI or ML?

I am open to AI/ML Engineering, Data Science, and Python roles, plus research collaborations and consulting. New York based, shipping worldwide.