At Mindrift, I evaluate LLM outputs, prepare training data, and provide structured feedback to improve generative AI models. I use Python and ML knowledge to identify edge cases, hallucinations, and reasoning errors.
At Netsol Technologies, I built LLM features for financial document extraction and summarization, and implemented LangChain- and RAG-based question-answering systems. I also integrated GPT and Gemini into production-style tools.
I built a voice-driven patient registration system with a FastAPI backend, VAPI, Supabase/PostgreSQL, Groq/Llama, and ElevenLabs; it achieved approximately 465ms end-to-end latency. My projects also include computer vision, supply chain forecasting, and an explainable customer churn dashboard.
