The Problem
Non-performing assets (NPAs) are an existential challenge for Indian banks. Trillions of rupees are locked in bad loans — money that can’t be re-lent, can’t fuel growth, and erodes profitability every quarter. Traditional collection processes are slow, manual, and expensive. Medius AI was built to change that.
System Architecture — Medius AI
Input
Bank / NBFC Data
- ›Loan accounts
- ›Borrower history
- ›Payment records
- ›NPA portfolio
Medius AI Backend
Python / Django · AWS · ML Integration
ML Risk Scoring
predict recovery likelihood
AI Decisioning
optimal contact strategy
Automated Workflows
multi-channel engagement
REST API Layer
bank system integration
Outcomes
Recovery Results
- ✓Reduced NPA ratios
- ✓Recovered capital
- ✓Faster loan resolution
- ✓Regulatory compliance
Live in production with Indian banks and NBFCs
What I Built
As Backend Team Lead, I owned the backend end to end — this wasn’t a supporting role.
System architecture: Designed and implemented the core Python/Django system from the ground up — data models, API contracts, and service boundaries. Built it to handle the scale and reliability requirements that financial institutions demand.
AWS infrastructure: Set up and maintained the cloud infrastructure — EC2, RDS, S3, and the networking layer. Designed for high availability and the security posture banks require.
ML integration: Wired the machine-learning scoring and decisioning models into the production API layer. The ML models output risk scores and recommended actions; the backend turns those outputs into real workflow triggers and multi-channel borrower engagement sequences.
API design: Designed clean REST APIs consumed by the frontend application and integrated with bank data feeds. Emphasis on reliability and clear error handling — financial data has no tolerance for ambiguity.
Documentation and engineering practice: Wrote technical documentation covering the system’s architecture, integration contracts, and deployment procedures. Drove code reviews, shaped engineering standards, and mentored junior engineers across the full SDLC.
Why It Matters
This is a production system trusted by financial institutions with real money, real borrowers, and real regulatory exposure. Banks don’t give that trust to prototype-quality work or to contractors who aren’t deeply embedded in the codebase.
Owning the backend of a live fintech platform — from the API layer to the infrastructure to the ML integration — is the clearest signal of what I can be relied on to do. If you need a senior engineer who can own a system, not just contribute to one, this is the reference.
Tech Stack
Python · Django · REST APIs · Third-party API Integration · AWS (EC2, RDS, S3) · PostgreSQL · SQL & NoSQL · ML model integration · Docker · Git