Backend Developer · DevOps Practitioner · AI/ML Integration Engineer

I build production-grade backend systems — from REST APIs and microservices to AI-powered automation. My focus is on clean architecture, real deployments, and integrating ML models into systems that actually run in production.

Backend: Node.js · Express.js · Python (FastAPI) · REST APIs · Microservices · JWT Auth
Databases: PostgreSQL · MySQL · MongoDB · Supabase
AI/ML: InsightFace · OpenCV · Gemini API · Face Recognition · NL2SQL
DevOps & Cloud: Docker · GitHub Actions · CI/CD · Linux (Ubuntu) · Caddy · VPS · AWS
Frontend: React.js · TypeScript · JavaScript · HTML5 · CSS3
Tools: Git · Postman · Power BI · n8n

Stack: Node.js · Express · React · PostgreSQL · Docker · GitHub Actions · Ubuntu VPS · Caddy

A production school ERP built as a freelance engagement. Handles student enrollment, billing, library, inventory, certificate generation, and reporting — deployed on a Hostinger VPS with Docker and Caddy reverse proxy.

• Modular backend with JWT auth and role-based access control across admin, teacher, and staff roles
• PostgreSQL database (Hostinger-managed) with optimized schema for multi-entity school data
• CI/CD pipeline via GitHub Actions — push to main triggers a Docker build and zero-downtime deploy
• Impact: Reduced administrative workload by 50%, improved data accuracy across departments

Stack: Python · FastAPI · InsightFace · OpenCV

A standalone microservice for face embedding generation and real-time identity matching. Integrated into Schoolix via REST — the Node.js backend delegates all biometric operations to this service.

• Generates and stores facial embeddings on enrollment; matches against stored vectors at attendance time
• Sub-second response times with 96% recognition accuracy in production conditions
• Fully decoupled — can be swapped or updated without touching the core ERP backend
• Impact: Replaced manual roll calls; increased attendance accuracy by 70%

Stack: Node.js · Gemini API (gemini-3.1-flash-lite-preview) · PostgreSQL · REST

A natural language interface layered over the Schoolix PostgreSQL database. Converts plain-text admin queries into valid SQL using Gemini with table relationship lineage rules and keyword-based intent resolution — no vector embeddings required.

• Schema-aware prompt construction with explicit table relationship lineage rules passed as context
• Keyword routing determines query intent before hitting the LLM — reduces hallucination on structured data
• Role-aware: responses are scoped to what the authenticated user's role can access
• Impact: Automates 80% of repetitive admin queries without any SQL knowledge required

Stack: React · TypeScript · Vite · Supabase · PostgreSQL

A general-purpose billing and patient management system — any doctor can register and start using it immediately. Handles the full billing lifecycle from patient registration to payment settlement.

• Custom charge line items with flexible billing form supporting variable pricing per consultation
• Rich-text patient notes persisted per visit record
• Installment payment flows with partial payment tracking and outstanding balance views
• Multi-doctor auth with role separation — each doctor sees only their own patient records
• Live: https://billing-system-pi-liart.vercel.app

Backend-focused engineer with hands-on DevOps and AI integration experience. I've deployed containerized applications on Linux VPS (Docker + Caddy), built ML microservices in production, and integrated LLMs into real-world data workflows. Currently targeting backend and AI integration roles at 6+ LPA.

• 🌐 Portfolio: https://portfolio-v5-vs.vercel.app/
• 💼 LinkedIn: https://www.linkedin.com/in/yash-dafade-992ab2209/
• 📧 Email: yashdafade93@gmail.com

Open to Backend, DevOps, and AI-Integrated Engineering roles.