📍 Punjab → Google, India | 🤖 AI/ML Engineer & Agentic AI Builder | 🎓 Educator · FreeBirds Crew

From manual tester at TCS to AI/ML Engineer at Google — no tier-1 college, no shortcuts. I build autonomous AI systems by day and teach thousands how to do it for free on YouTube by night.

• ⏳ Chronos — Autonomous AI Research Harvester, a zero-cost, privacy-first research pipeline that automatically aggregates papers from arXiv, Hugging Face, OpenAI, Anthropic, Google, and DeepMind, evaluates them using a local Ollama model (Qwen2.5), and tracks associated GitHub code links and star counts — all written seamlessly into a live Google Sheets database via a background macOS daemon.

• 🔁 Lookback — Multi-Agent Interpersonal Simulator, an advanced relationship simulation system that models, analyzes, and visualizes human communication dynamics using local LLMs (via Ollama) mapped to clinical psychology frameworks like Attachment Theory and the Big Five personality traits — all streamed live to a React interface using FastAPI and Server-Sent Events (SSE) streaming.

• 🎯 OpportunityOS — Autonomous AI Content Engine, an end-to-end autonomous research pipeline that ingests papers from arXiv, Hugging Face, and Anthropic, grades them with Gemini 3 Flash, and auto-generates viral-ready social posts for X, LinkedIn, and Instagram — all streamed live via a glassmorphism React dashboard.

• ⏳ KAAL — KAAL (Knowledge Augmentation & Autonomous Logic) is a sovereign AI operating layer for macOS. It is designed as a cognitive exoskeleton—an intelligence layer that amplifies your mind, knows your context, and acts autonomously on your behalf.

• 🧠 Synapse — Enterprise Agentic RAG System. An enterprise-grade RAG monorepo powered by Google Gemini 3 featuring self-healing retrieval (CRAG), adaptive query routing, hybrid Dense+Sparse search, and multi-layer security guards (input/output/PII) — built for citation-accurate answers at million-document scale.

• 🤖 FreeBirdsCrew WhatsApp AI Bot — LLM Wiki-Powered WhatsApp Assistant. A smart WhatsApp bot built on Karpathy's LLM Wiki Architecture using Gemini 3.0 Flash — routes incoming messages to the right resources (YouTube playlists, GitHub repos, Topmate sessions) via intent classification, with a built-in React dashboard and keyword fallback.

• 🚀 AndrejKarpathy 2026 Focus Projects — A curated 7-project roadmap (inspired by Andrej Karpathy) covering Context Engineering, Tool Design, Agent Orchestration, Eval Discipline, and Reliability Engineering — progressing from PR review agents to a full autonomous DevOps incident response system.

• 🥭 MangoOS — Agentic AI Operating System — Production-grade, fully self-hosted AI OS; multi-agent orchestration via LangGraph, Docker-sandboxed execution, Redis + ChromaDB memory, RBAC guardrails, and a React/Electron desktop UI — all running locally on Ubuntu with Ollama.

• 🦾 Awesome AI Agents — 17+ autonomous agents from scratch; the most comprehensive free Agentic AI course on GitHub. 157★

  • 🏥 MedGemma Analyzer — AI radiologist: high-fidelity medical imaging interpretation using MedGemma 1.5 + Streamlit
  • 📈 AlphaAdvisor — Autonomous investment strategist; real-time financial analysis + stock recommendations via Gemini + Yahoo Finance
  • ⚖️ LegalCounsel AI — Multi-agent legal team for contract review & research; Gemini + Agno + ChromaDB
  • 💻 DevForge Agent — End-to-end SDE assistant that plans, writes, and tests code autonomously via OpenAI Swarm
  • 🩺 VisionDx Agent — Vision-based radiology assistant powered by Gemini multimodal; upload scans, get analysis
  • ✈️ NomadAI Planner — Autonomous travel planner; custom itineraries, flights & hotels via SerpApi
  • 🧠 DeepSeek RAG Agent — Fully local reasoning agent using DeepSeek-R1 + Ollama; no cloud, no cost
  • 🔮 Gemma3 RAG Agent — Same offline-first architecture with Google Gemma 3; privacy-first local LLM reasoning
  • 🎥 VideoMind Agent — Intelligent video analysis for YouTube URLs & uploads; Gemini + OpenCV
  • 🏋️ FitCoach AI — Personal nutrition & workout coordinator built on a multi-agent Gemini pipeline
  • 🧬 MemoryMind Agent — Chatbot with persistent cross-session user memory via Mem0 + Gemini
  • 🔬 ResearchSwarm — Deep-dive autonomous research using OpenAI Swarm multi-agent coordination
  • 🕸️ LeadHarvest Agent — Lead generation from Quora, Amazon & job boards using Firecrawl + Composio
  • 📄 AgenticRAG Flash — Advanced PDF + web-search RAG using Gemini Flash Thinking + Exa
  • 🏭 ADK SaaS Boilerplate — Production-ready SaaS agent scaffold on Google ADK; clone and ship
  • 🛠️ ADK Toolkit — Beginner-to-advanced Google ADK implementations collected in one place
  • 🌦️ InfoStream Agent — Real-time weather + news delivery agent using DuckDuckGo search tool

• 🌌 Sams — Personal OS Agent — Karpathy-style autonomous research loop; set a goal overnight, wake up to structured Markdown reports. Docker-sandboxed, MCP-ready, adversarial AutoReason engine inside.

• 🔄 NeuralForge — Self-Extending Agent — Agent that writes and registers its own new skills at runtime on Google ADK; full-stack (Python backend + JS frontend + dev assistant app).

• 📊 StockMarket ADK — Production multi-agent trading system for NSE/Nifty 50; Analyst + Risk + Execution agents, SEBI 2026-compliant, Zerodha Kite integration, glassmorphism dashboard.

• 🎮 ARISE — Solo Leveling AI macOS — Gamified autonomous life-OS for macOS inspired by the Solo Leveling manga; Swift menu-bar agent powered by Gemini.

• 📚 RAG from Scratch — LLMs — Complete hands-on RAG curriculum: HuggingFace + LangChain + Python, zero to production.
  • 🔗 Pinecone RAG Pipeline — Build end-to-end: chunking → embeddings → Pinecone vector store → retrieval → generation
  • 🧪 RAG Evaluation Suite — Measure your RAG pipeline: precision, recall, faithfulness, and context relevance

• 🏆 75DayHard GenAI Challenge — A 75-day public learning gauntlet to master LLMs; all source code, Kaggle notebooks, and daily takeaways. 49★
  • 🤗 HuggingFace Streamlit App — Deploy open-source HuggingFace LLMs with a Streamlit UI in under 30 minutes
  • 🖼️ ChatWithImages — Multimodal chatbot: upload any image, ask natural language questions about it
  • 🔤 NextWord Predictor — Next-word prediction model built from scratch with deep learning fundamentals
  • 🎬 TextToVideo PyTorch — Generate short video clips from text prompts using PyTorch diffusion models
  • 🦙 LLaMA 2 QLoRA Fine-Tune — Fine-tune Meta LLaMA 2 on custom data with QLoRA; consumer GPU friendly
  • 🔮 Dolly v2 LoRA Fine-Tune — Instruction-tune Databricks Dolly v2 with LoRA adapters from scratch
  • 🌐 Phi-2 QLoRA Fine-Tune — Fine-tune Microsoft Phi-2 with QLoRA; small model, surprisingly powerful results
  • 🟢 LLaMA 3 ORPO + PEFT — LLaMA 3 preference alignment via ORPO + LoRA + PEFT; state-of-the-art alignment technique
  • 🌸 Gemma Fine-Tune — Domain-specific fine-tuning of Google Gemma 1.5B on custom datasets
  • 💬 Gemma Chatbot — Conversational chatbot built on Gemma with full multi-turn chat history
  • ⚙️ LLM Quantization Complete Guide — GPTQ vs AWQ vs bitsandbytes: practical comparison with benchmarks
  • 📐 Quantization Deep Dive — Binary vs integer quantization explained with hands-on experiments
  • 🌐 Wikipedia GPT-2 LM — Train a language model on Wikipedia using GPT-2; understand pre-training from scratch
  • 🗄️ Pinecone RAG (75-Day) — Production-grade RAG with Pinecone vector DB and an LLM; challenge Day final project

• 💰 LLM RAG Finance Use Cases — Real-world GenAI + RAG applications in Finance; theory + production-grade implementations for the domain.
• 📉 Financial Analysis Python — Financial data analysis toolkit: stock screening, portfolio metrics, and visualization in Python.

• 🏏 IPL Data Analysis — Deep EDA on IPL dataset: player stats, team trends, data profiling + visualization.
• 🎬 Netflix Data Analysis — Age-rating trends, top genres, and content patterns in Netflix's 2021 catalog.
• 🦠 Covid-19 Data Analysis — Impact analysis and solution mapping from global pandemic datasets.
• 🤖 ChatBot TensorFlow NLP — Contextual chatbot using TensorFlow + NLP that understands intent and context of human language.
• 🌐 Programmer Portfolio — Personal developer portfolio: projects, skills, and source code all in one place.

• Building at Google — designing and shipping scalable AI/ML systems in production
• Teaching on YouTube (FreeBirds Crew) — weekly videos on Agentic AI, LLMs, RAG, and research paper explainers; always free, always project-based
• Exploring self-extending agents — systems that write and register their own new tools at runtime
• 2026 Mission — contribute meaningfully to the Agentic AI open-source ecosystem and help engineers across India break into AI without needing a tier-1 degree

• The DeepSeek OCR Paper Explained | AI Can Now See Text Instead of Reading
• Build 9 End-to-End GenAI Projects 2026 | RAG + Agentic AI + LLMs | Scalable System Design
• TCS to Google | From Manual Tester to ML Engineer | Punjab to FAANG | Tier 3 to GenAI

• 🦾 Agentic AI — 14 Complete Project Tutorials
• 🗂️ Learn RAG from Scratch with Projects
• 🏋️ GenAI 75 Hard Challenge + Projects
• 📐 GenAI/LLM Research Papers Explained with Maths

• AI/ML Engineer at Google — the destination after years of relentless self-investment from a tier-3 college in Punjab
• 157+ stars on Awesome AI Agents — the community's go-to free Agentic AI curriculum
• FreeBirds Crew — built an active community of AI learners getting into the field without expensive courses
• Shipped 30+ open-source AI projects spanning Agentic AI, RAG, LLM fine-tuning, Finance AI, and Data Science

"Your background doesn't decide your destination — your consistency does." I build open-source AI tools and teach for free because I needed this content when I was starting out and it didn't exist.

• Started career as a manual tester at TCS — zero ML knowledge, zero connections
• Cracked Google without a tier-1 degree — the long way, the real way
• Runs multiple LLM instances in parallel for research, experiments, and content
• Unwinds by drawing, playing guitar, and writing songs and poetry
• Currently obsessed with self-extending agents that write and register their own tools
• Believes the best way to master AI is to ship a project and teach it publicly