I've deployed production AI systems at Join Venture AI for international healthcare, enterprise automation, conversational AI, and construction intelligence clients. My work includes RAG pipelines, vector search architectures, multimodal OCR, LLM reasoning, and scalable FastAPI inference services.
For Cosmaworks AI, I delivered Business CS Bot and Friend Bot, serving 5,050 personas through fine-tuned Qwen models, RAG, FAISS vector search, conversational state machines, guardrails, and isolated SQLite memory. I reduced fine-tuning time from approximately 73 hours to approximately 2 hours and deployed FastAPI and vLLM services on RunPod A6000 infrastructure.
I've also built clinical documentation and medical-assistant platforms using Claude, GPT-4o, Pinecone, and vision APIs, construction quantity-takeoff automation using OCR and Gemini/OpenAI, and a voice-enabled AI receptionist SaaS with ElevenLabs and Twilio.
