Software Engineer at Oracle, working on CPQ backend systems and automation.
I build applied AI systems where the model is one component in a larger product: retrieval, evaluation, business rules, APIs, persistence, and deployment all have to hold together.
| AI systems | Backend | Frontend | Data and infra |
|---|---|---|---|
| LangChain, LangGraph, RAG, AI agents, LLM evaluation, Pinecone, n8n | Python, Django, DRF, FastAPI, Flask, REST APIs | React, TypeScript, JavaScript, Tailwind CSS | PostgreSQL, MongoDB, SQLite, Docker, Render, Vercel, Cloudinary |
| Signal | Evidence |
|---|---|
| Applied AI | RAG evaluators, agentic CPQ workflows, resume-screening agents, market-signal automation |
| SlotBuddy | Hackathon runner-up, best project in n8n category, and commercial product used by 3 clients |
| Problem solving | 4-star CodeChef, rating 1958 |
| Education | B.E. ETCE, Jadavpur University |
|
LLMOps dashboard for choosing the best model + RAG configuration across quality, citations, latency, cost, and failure risk. Built with: Python, FastAPI, Pinecone, BM25, reranking, RAG evaluation, OpenAI/Groq/Gemini, Vercel |
Agentic CPQ workflow that turns natural-language sales requests into editable SaaS quotes with deterministic pricing, RAG policy retrieval, and PDF generation. Built with: Django, DRF, LangGraph, React, TypeScript, Pinecone, ReportLab, PostgreSQL |
|
LangGraph multi-agent workflow that compares resumes against job descriptions, flags risks, applies weighted scoring, and ranks candidates with transparent reasoning. Built with: Streamlit, LangChain, LangGraph, OpenAI, Python |
AI pipeline that mines Reddit discussions, clusters market pain points, scores startup ideas, and produces an editable newsletter draft. Built with: FastAPI, Python, SQLite, OpenAI, Gmail drafts, SMTP/Resend integrations |
|
Real-time collaborative code editor for interviews with shared rooms, live editing, and a CodeMirror-based coding surface. Built with: React, Express, Socket.IO, MongoDB, CodeMirror |
Django package for serving static files through Cloudinary CDN, reducing server load and improving static asset delivery. Built with: Python, Django, Cloudinary, PyPI packaging |
SlotBuddy automates appointments, reminders, calendar updates, and CRM workflows for service businesses.
- Runner-up at ElevenLabs Hyderabad Hackathon
- Best project in the n8n category
- Commercially used by 3 clients
- Public reference: Hackathon Demo
- LLMs are useful; deterministic backend logic still owns prices, permissions, totals, persistence, and auditability.
- RAG work needs evaluation, not vibes: expected answers, citation checks, latency, cost, failure categories, and traces.
- A good side project should feel like a small product: usable UI, clear API boundaries, deployment path, docs, and fallback behavior.