Skip to content

← 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.
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.