AI Product Engineer — I build things that work in production, not just in demos.
AI Product Engineer at Ruby CRM / CleverFlow (Dubai). I build production AI systems — RAG pipelines, document intelligence, computer vision, LLM fine-tuning. 20,000+ req/s. 10,000+ documents/month. 45+ features shipped. 99.5% uptime.
Previously: fine-tuned medical LLMs, published 5 models on Hugging Face, won Vadodara Startup Festival (250+ startups), hold a design patent for medical IoT.
|
Optimize Claude Code token usage. 22 skills, 8 hooks, 6 rules. One npm i -g claude-code-optimizer
|
Subflo — In Progress AI-powered subscription tracker. Scans Gmail for payment receipts, auto-detects recurring subscriptions using any LLM (Ollama, OpenRouter, Groq, OpenAI). No bank API. No card linking. Just Gmail + AI.
|
| Model | Downloads | What It Does |
|---|---|---|
| MedGenius_LLaMA-3.2B | 1,200+ | Medical LLM — 89% accuracy, +86% ROUGE-1 over base |
| MedGenius_LLaMA-3.2B-Q4_K_M-GGUF | 400+ | Quantized version for local deployment |
| Medical Intelligence Dataset | 800+ | 40,443 medical dialogues — open source |
| System | Scale | Tech |
|---|---|---|
| Document Intelligence API | 10,000+ docs/month | FastAPI, OpenCV, LLM Vision, OCR |
| Email Intelligence (RAG) | Production pipeline | Qdrant, SentenceTransformers, Redis |
| Floor Plan Inspector | CV compliance tool | Computer Vision, YOLO |
| WhatsApp AI Agents | Customer automation | LLMs, WAHA, n8n workflows |
| Vector Search Optimization | 60% latency reduction | Qdrant, hybrid search, RRF |
| Redis Caching Layer | 40% API cost reduction | Redis, FastAPI |
AI/ML PyTorch · Transformers · LangChain · LlamaIndex · LoRA/QLoRA · GGUF
LLMs Fine-tuning · RAG · Embeddings · Vector Search · Qdrant · Faiss
CV OpenCV · YOLO · Detectron2 · OCR
Backend FastAPI · Node.js · Express · PostgreSQL · Redis · MongoDB
Frontend Next.js · React · TypeScript · Tailwind
DevOps Docker · GitHub Actions · CI/CD · AWS S3 · Vercel
- Stop Wasting Tokens: How I Cut My Claude Code Costs by 67% — Medium
- Medical AI: Fine-Tuning LLaMA for Healthcare — Blog
- Building RAG Systems That Actually Work in Production — Blog
"Two hours of scrolling won't change your life. Two hours of building might."