██████╗ ██████╗ ██╗ ██╗ █████╗ ███╗ ██╗ ██████╗ █████╗ ████████╗██╗██╗
██╔══██╗██╔═══██╗██║ ██║██╔══██╗████╗ ██║ ██╔══██╗██╔══██╗╚══██╔══╝██║██║
██████╔╝██║ ██║███████║███████║██╔██╗ ██║ ██████╔╝███████║ ██║ ██║██║
██╔══██╗██║ ██║██╔══██║██╔══██║██║╚██╗██║ ██╔═══╝ ██╔══██║ ██║ ██║██║
██║ ██║╚██████╔╝██║ ██║██║ ██║██║ ╚████║ ██║ ██║ ██║ ██║ ██║███████╗
╚═╝ ╚═╝ ╚═════╝ ╚═╝ ╚═╝╚═╝ ╚═╝╚═╝ ╚═══╝ ╚═╝ ╚═╝ ╚═╝ ╚═╝ ╚═╝╚══════╝
class RohanPatil:
def __init__(self):
self.languages = ["Python", "SQL", "JavaScript", "C"]
self.ml_stack = ["Scikit-learn", "NumPy", "Pandas", "Matplotlib", "Seaborn"]
self.dl_stack = ["Neural Networks", "CNNs", "Transformers", "Fine-tuning"]
self.llm_stack = ["Prompt Engineering", "RAG", "LLM APIs", "Hugging Face", "Transformers"]
self.cv_stack = ["OpenCV", "MediaPipe", "Real-time Detection", "Landmark Tracking"]
self.backend = ["FastAPI", "Flask", "Streamlit", "REST APIs"]
self.open_source = ["Mozilla Firefox Relay", "Google DeepMind · Gemma", "Gemini CLI", "OpenClaw"]
self.frontend = ["HTML", "CSS", "JavaScript"]
self.currently = "shipping ML models to production + levelling up on RAG & Vector DBs"
def run(self):
while True:
self.learn() → self.build() → self.contribute() → self.ship()📦 Mozilla · Firefox Relay → Frontend accessibility · WCAG AA contrast compliance
🧠 Google DeepMind · Gemma → Dataset observability · DPO preprocessing pipeline
💎 Gemini CLI → Contributed to Google's open-source Gemini terminal client
⚙️ OpenClaw → AI API cost analysis utility · error handling · docs
┌─────────────────────────────────────────────────────────────┐
│ AI / ML / DL │
│ ├── Supervised · Unsupervised · Regression · Classification │
│ ├── Neural Networks · CNNs · Transfer Learning │
│ └── Model training · Evaluation · Deployment │
│ │
│ LLM & Generative AI │
│ ├── Prompt Engineering · Chain-of-Thought · Few-shot │
│ ├── Hugging Face Transformers · LLM API integration │
│ ├── RAG · Vector Databases · Embeddings │
│ └── Local model inference · Context window management │
│ │
│ Computer Vision │
│ ├── OpenCV · MediaPipe · Real-time webcam pipelines │
│ └── Hand landmark detection · Gesture recognition │
│ │
│ Backend & Deployment │
│ ├── FastAPI · Flask · Streamlit │
│ └── REST APIs · Git · CI/CD basics │
└─────────────────────────────────────────────────────────────┘
| # | Project | Stack | TL;DR |
|---|---|---|---|
| 01 | Gesture-Controlled Hill Climb | OpenCV MediaPipe PyAutoGUI |
Real-time hand landmark → keypress. Zero hardware. |
| 02 | Multi-Agent AI Research System | Python LLMs Agent Frameworks |
Collaborative AI agents for automated research |
| 03 | Spam Detector | Scikit-learn NLP TF-IDF |
Full ML pipeline: data → features → model → prediction |
| 04 | House Price Predictor | Pandas Regression EDA |
Feature engineering + regression from scratch |
| 05 | HackMET 2026 Site | JavaScript Frontend |
Production site — 500+ participants, ₹5L prize pool |
✅ Machine Learning Specialization — Andrew Ng / Stanford / Coursera
✅ Generative AI — Google Cloud
✅ Large Language Models — Google Cloud
✅ Prompt Engineering for GenAI — LinkedIn Learning
✅ Responsible AI — Google Cloud Skills Boost
✅ Generative AI with Adobe — Adobe→ FastAPI + Streamlit → deploying ML models as real web apps
→ RAG Pipelines → retrieval-augmented generation from scratch
→ Vector Databases → FAISS · Chroma · embeddings
→ Transformers → fine-tuning open-source LLMs (Gemma, LLaMA)
→ Open Source → moving from bug fixes → feature-level PRs
python -c "print('learn → build → contribute → ship')"