Hire Imran F. - Fine-Tuning Engineer

* Built Nexa, a production AI health companion with a LangGraph multi-agent chatbot routing user intents across biomarker goal-setting, deep research, general chat, and pantry discovery workflows. * Implemented real-time WebRTC voice chat with safety guardrails for AI wellness conversations. * Developed an OpenAI-powered blood-test PDF parser that extracts biomarkers and supports personalized wellness goal generation. * Engineered a hybrid RAG pipeline using OpenAI embeddings, BM25 keyword search, reciprocal rank fusion, and biomarker knowledge documents. * Productionized FastAPI backend services with Redis-backed rate limiting, Supabase persistence, pytest coverage, and production-mode safety checks.

* Worked with OSF HealthCare on DestinAI, an AI decision-support proposal for interfacility patient transfer routing across a multi-hospital healthcare network. * Expanded the original proposal into a feature-level system design, defining required workflows for hospital transfer recommendations. * Architected key features including rule-based screening, LLM-assisted recommendation logic, and human-in-the-loop review. * Designed wireframes and workflow screens that were presented at OSF Innovation 2026. * Developed a validation framework to support the safety, reliability, and clinical justification of AI-assisted hospital transfer recommendations. * Worked on implementation planning for a rule-based + LLM recommendation system using patient, facility, and operational constraints.

* Benchmarked six open-weight LLMs including DeepSeek, LLaMA-2, Nous-Hermes-2, Phi-3, Gemma, and Mistral for fake news classification. * Compared zero-shot prompting, QLoRA fine-tuning, classical RAG, RAPTOR retrieval, and neurosymbolic reasoning approaches. * Fine-tuned models using QLoRA with 4-bit quantization and evaluated performance using accuracy, precision, recall, and F1-score. * Built retrieval pipelines using FAISS, vector embeddings, and prompt-based evaluation workflows. * Published the comparative LLM fake news detection study at IEEE ICSC 2026.