Python backend developer and AI engineer building systems end-to-end.

• REST API design, database integration, concurrency-safe backends, and role-based access control
• Agentic AI / MCP / LangGraph, from protocol-level tool-calling to human-in-the-loop workflows
• Test-driven: every flagship project below ships with a real test suite, not a demo script

• Bengaluru, India
• I like building things that actually run: APIs, agentic pipelines, and concurrency-safe backends, not just notebooks
• Comfortable going from model training and evaluation through to deployment and wiring the result into a real backend
• Curious about secure system design: how systems fail in production, and how authorization and architecture choices prevent that

• REST APIs with FastAPI/Django, JWT auth, and relational data modeling
• Concurrency-safe backends: row-level locking, cache invalidation, idempotent writes
• Agentic AI systems: MCP servers/clients, LangGraph tool-calling agents, human-in-the-loop workflows
• Backend systems with clean structure, auditable data, and role-based access

Django DRF PostgreSQL Redis Celery JWT

• Load tested with Locust: without row-level locking, 40 concurrent users silently lost 16 units of stock with zero HTTP errors; select_for_update() closes the race to zero discrepancy
• Admin / Manager / Employee RBAC with JWT access and DB-blacklisted refresh tokens, rate limited per scope; append-only stock ledger enforced twice (no write route registered, and Django Admin permissions hard-disabled)
• Redis-cached reports with version-counter invalidation, cache hits collapse 9 queries down to 1 (proven via query-count assertions, not just timing)
• Celery background jobs and Beat schedule for async invoices, low-stock alerts, and nightly ledger reconciliation; deployed on AWS behind an Application Load Balancer

GitHub · Live Demo

FastAPI Async SQLAlchemy JWT ImageKit

• Async FastAPI backend with JWT auth via fastapi-users: register, login, email verification, forgot/reset-password
• Media streamed to a temp file then pushed to the ImageKit CDN, with UUID-keyed async SQLAlchemy models
• Ownership-based authorization returns 403 on delete attempts by non-owners
• REST API and static frontend served from a single FastAPI process, with zero CORS overhead

GitHub

Django DRF PostgreSQL JWT pytest

• Django REST Framework backend for multi-account tracking, categorized transactions, monthly budgets, and spending reports
• JWT auth (djangorestframework-simplejwt) with per-user data ownership enforced end-to-end on every endpoint
• Versioned API under /api/v1/, live-computed account balances and budget spent/remaining figures, no stale duplicated values
• pytest + pytest-django coverage suite; interactive Swagger/ReDoc docs via drf-spectacular

GitHub

MCP FastMCP Groq FastAPI Pytest

• Real FastMCP server exposing Tools, Resources, and a Prompt as its own separate process; the agent reaches every one of them only through an MCP ClientSession over stdio, never by importing a Python function directly
• Live "what the client discovered" panel and a per-turn call trace show the actual list_tools / list_resources / list_prompts results and call_tool arguments, not simulated for the UI
• Swapping in a real third-party MCP server (filesystem, GitHub, Slack) only means changing the StdioServerParameters, and nothing about the agent loop changes, proving the protocol boundary actually holds
• Test suite spins up the real MCP server subprocess and drives it through the full protocol (discovery, tool calls, resource reads, prompt templates) with no API key or network needed

GitHub

LangGraph FastAPI SQLite Groq Pytest

• LangGraph agent adds, lists, updates, completes, and deletes tasks, and pauses mid-run for human confirmation before any delete via interrupt() / Command(resume=...)
• Conversation state persisted across requests with a SQLite checkpointer, so context survives a page refresh, not just an in-memory session
• Test suite swaps in a scripted FakeModel to exercise the real graph, real tools, and real interrupt/resume flow with no API key or network call
• FastAPI backend with a raw request/response inspector in the UI, showing exactly what the agent sent and received on each turn

GitHub

• drkrillo/good-first-issues – CI smoke test for the action itself (#172)
• Ayorinha/ayorai-vision-intelligence – Restored corrupted CI workflow YAML (#26)
• yunaremaia/gfi – LICENSE, Dockerfile & test coverage (81%→93%) (#52)

Backend Development · Python, FastAPI, REST API Design, SQLAlchemy & Alembic, JWT Auth & RBAC, Async Programming, Pytest, Git & GitHub

Agentic AI · Generative AI, Prompt Engineering, Tool Calling, RAG, LangChain, LangGraph, AI Agents, Model Context Protocol (MCP)

Full breakdown with the complete tag list is on the live portfolio site.

• AWS Solutions Architect Associate (SAA-C03) – Packt (Jul 2026)
• IBM RAG and Agentic AI Professional Certificate · IBM (Aug–Sep 2026)
• Artificial Intelligence (Credit Course) · SmartBridge × Google for Developers (May–Jun 2025)
• AI Fluency: Framework & Foundations · Anthropic
• Introduction to Model Context Protocol · Anthropic
• Model Context Protocol: Advanced Topics · Anthropic
• Claude Code in Action · Anthropic
• Problem Solving, Java, Python, SQL (5-star) · HackerRank

🔗 LinkedIn 🔗 HackerRank 🔗 LeetCode 🔗 Hugging Face