← Aman Kumar Limited · RegisterNotes to the accounts · MESH
Featured · agent · 2025 · Shipped
MSP Intelligence Mesh
A federated-learning network where IT service providers share threat and client-health signals without sharing raw data. Built for a worldwide hackathon.
- Role
- Hackathon build — architecture, agents, backend and frontend
- Lead
- Outcome
- A federated-learning network where IT service providers share threat and client-health signals without sharing raw data. Built for a worldwide hackathon.
- Links
- Code ↗
- Stack
- Python · FastAPI · React · AWS serverless · Federated learning · Differential privacy · Docker
Problem
Managed IT service providers each see a small slice of threats and client churn. Pooling data would help everyone, but nobody can share client data.
What I built
- A federated-learning mesh. Each provider trains locally; only model updates are shared, protected with differential privacy.
- Ten specialised AI agents — threat intelligence, client health, revenue optimisation, partner matching, market intelligence and more — working over the shared signals.
- A full stack on AWS serverless with a FastAPI backend, a React dashboard and Docker for local runs.
Result
Built end to end for a worldwide hackathon (growth and financial-improvement theme), November 2025.
Risk factors
- Hackathon prototype: the accuracy and revenue figures in the repo README are demo projections, not production results.
This is a portfolio, not an offer of securities. No shares exist. Figures are about shipped work.