hi, I’m Ravish 👋

I ask “why is a human doing this?” a lot.

AI Engineer focused on turning cutting-edge AI into production software. I build LLM applications, RAG pipelines, AI agents, and document automation systems with Python, FastAPI, React, Next.js, Azure, and Docker.

Currently AI/ML Engineer at DataFactZ · New Delhi, India.

Contact GitHub ↗ no template, all shipped ✨
70%less manual data entry from OCR automation
60%operating cost savings on document processing
40–45%faster legacy BI migration with LLM tooling
80–90%answer accuracy, citation-backed RAG chatbot
Anthropic · Certification

Claude Certified Architect - Foundations

Anthropic’s credential for architecting Claude-powered systems.

✓ Verify on Credly ↗
where I’ve worked

Work experience

DataFactZ

Hyderabad
  • AI/ML EngineerOct 2024 – Present
  • AI/ML Engineer · ApprenticeshipAug 2024 – Oct 2024 · 3 mos
PythonFastAPIReact / TypeScriptAzure Document IntelligenceAzure AI SearchAzure OpenAIAzure AI FoundryPower BI REST APIDockerRedis
and after hours…

Personal projects

Restaurant Management System

Dec 2025 – Present · Founder

Built it, sold it, runs it - a full-stack ordering platform live across 3 real restaurants.

  • QR-code dine-in ordering, a real-time Kitchen Display System, and live counter/customer display boards - all synced over Socket.IO (room-based order, kitchen-bump & menu events).
  • Custom RBAC engine - 97 permissions across 12 categories with wildcard grants (*, menu.*) driving 11 built-in roles from super-admin to guest.
  • Configurable discount & coupon engine and 8 payment methods (UPI, cards, wallets, pay-at-counter) via Razorpay, with full refund / partial-refund handling.
  • 25-table Drizzle schema on libSQL/Turso - a 7-state order lifecycle with status-history audit trail, per-day order tokens, and an R2 image pipeline (sharp → WebP + blur placeholders).
Next.js 16React 19TypeScriptDrizzle ORMlibSQL / TursoSocket.IORazorpayNextAuthCloudflare R2
Live demo - divinefeast.ostepbig.shop ↗

Inbox CFO

Personal · Built solo in 48 hours

Your inbox, turned into a finance dashboard - the AI finance agent that catches the charge before it hits your card.

  • Turns Gmail (read-only) or uploaded .eml/.pdf into typed rows in one JSON-mode LLM call - flagged ₹13,992/yr in trials, price hikes & duplicate subs on a seeded inbox.
  • Proactive, not reactive: reads the warning email days before the charge and drafts the cancellation - caught an Audible auto-charge 2 days early.
  • Zero hallucinated totals - every number is deterministic SQL with a source_id per row; click any figure to open its source email.
  • Grounded chat agent answers spend questions via tool-calls over the tables (read + draft only, 5-loop cap) and draws its own sourced charts.
React / ViteTailwindRechartsFastAPISQLModelSQLiteGemini 2.5 / Llama 3.3Gmail API
Live demo - inboxcfo-frontend.onrender.com ↗ Source - github.com/ravish0409/InboxCFO ↗
my toolbox

Skills & stack

Agentic AI & GenAI

RAG pipelines AI agents Multi-agent orchestration Tool & MCP design Embeddings & vector retrieval Prompt engineering Structured output Production reliability

Backend & Languages

Python TypeScript FastAPI JavaScript C/C++ SQL · MySQL · SQLite Django Express REST APIs WebSockets / Socket.IO JWT / OAuth

Frontend

React Next.js Tailwind Streamlit

Cloud & DevOps

Azure Document Intelligence Docker Git / GitHub Cloudflare R2 Linux

Concepts

RBAC OCR pipelines Async & microservices Schema design & migration SSO / OAuth
receipts 📎

Credentials

Education

BCA, GGSIPU (Jagannath International Management School), 2024 - CGPA 8.2.

XII CBSE 84% · X CBSE 85%, Deepalaya School.

Achievements

Ranked 26th - CodeChef April 2022 Long Challenge.

AIR 523 - IPU entrance exam.

Ranked 153rd - HackerEarth October Circuits ’22.