AI engineer · Data scientist · New York
Eshaan Mathakari
I build AI systems for a living and keep building them after hours. This site collects the research, the side projects, the things that worked and the things that didn't.
nestAntelligence0 tripsslaya0 tripsHome agent fleet0 tripsGrow simulator0 tripsSwarm Oracle0 tripsHonest to a Fault0 tripsScoutDocs0 tripsVelo0 trips
Live: ants leave the nest, find a project and lay a trail home. Other ants follow it, so busy trails get busier. Hover a project to send ants its way.
- Based in
- New York City
- Day job
- AI Engineer at EXL
- Researching
- slaya and Antelligence
- Home lab
- Running since spring 2026
Selected work
All 50 projects- AntelligenceAnt-colony teamwork for swarms of tiny robots, tested in simulation and aimed at medicine.

- slayaSmall, fast models that answer yes/no and probability questions so a big AI doesn't have to.
- The home agent fleetAn always-on team of AI agents that runs on a Mac Studio in my apartment.
- What grows on my block?A simulator for which plants will actually grow where you live in New York.
- Swarm OracleSeveral AI agents forecast the same question, and the ones with better track records get more say.
- Honest to a FaultCan you make an AI model more honest by turning up its internal "honesty" signal?

- ScoutDocsGives AI coding assistants the documentation for the exact software version you are using.
- VeloTennis coaching feedback from a single phone video.
Professional work
All professional workBuilt during my professional work. The names are project code names, and client details stay private. Several are prototypes or proofs of concept built on synthetic data.
- Pitch CompassA prototype AI copilot that helps bank sales advisors understand business clients and choose next steps.
- Fine Print HawkA workspace that checks loan covenants against their source documents and flags every mismatch.
- Paper TrailFour AI agents run business due-diligence checks, while fixed rules decide every risk score.
- Rulebook RadarEight AI agents turn a regulatory document in any language into a reviewed change record.
- Ticket WhispererSix AI agents take an IT support incident from first report all the way to closure.
- Sparring PartnerAn AI red-team and blue-team platform that attacks chatbots, scores the results and proposes fixes.
- Ink SlingerAn AI assistant that redlines NDAs and purchase terms, citing a rule for every edit.
- Sticky FingersA shared hub where teammates publish reusable skills that AI coding agents can find and fetch.
- Inbox ButlerTwo automation pipelines that replaced manual inbox monitoring and data entry for order processing.
Right now
October 2026- slayaBuilding a better dataset for small decision models before training anything bigger.
- AntelligencePlanning the next colony experiment with Kashyap, built around the merge step.
- What grows on my block?Working out which data a plant simulator for New York actually needs.
- Day jobBuilding agentic AI systems at EXL.
Recently pushed
snapshot · 2026-09-17- solmonger/universal-docs-mcp
- eshaanmathakari/scoutdocs-mcp
- solmonger/thetalktherapy
- solmonger/antelligence-app
- solmonger/honest-to-a-fault
- eshaanmathakari/sholay
Public repositories only. Updates live from GitHub when it can.
Writing
All posts- What an ant colony taught us about teams of AI modelsWe put several different language models into one simulated colony to see if the mix would beat its best member. The answer pointed somewhere more useful.
- Six months of agents in my apartmentI've run a fleet of AI agents on a Mac Studio at home since spring. What held up, what didn't, and the failures that looked exactly like success.
- About ten dollarsMonths of building prediction-market bots earned roughly the price of a sandwich. Here is what the money couldn't measure.