At Hubblehox Technologies, I build production AI systems spanning YOLOv8/OCR extraction, RAG assistants, AI interview automation, proctoring, and Qwen fine-tuning. My computer-vision pipelines reached 95% extraction accuracy, improved detection precision by 15%, and helped reduce teacher hiring time by more than 50%.
I also evaluate Claude and Gemini outputs and coding agents for correctness, safety, instruction following, and code quality. I work hands-on with Python, Flask, FastAPI, vector retrieval, guardrails, LLMOps, and Google Cloud Run, scaling AI workflows to roughly 70–90 requests per second.
