Sree Harsha

I’m an AI/ML Engineer focused on generative AI systems across NLP and computer vision. I build and evaluate end-to-end pipelines—from data preprocessing and model training to deployment—aiming to improve performance and reliability.

In my internship and freelance work, I’ve developed real-time AI pipelines with PyTorch, OpenCV, and MediaPipe, and worked on 3D avatar generation using Gaussian Splatting and COLMAP. I also evaluate LLM outputs using rubric-based scoring, A/B ranking, and error analysis to identify hallucinations and reasoning issues.

My projects include a Local RAG Chat System using offline, privacy-preserving LLMs (Ollama) for better grounding and fewer hallucinations, an AI-powered cybersecurity scanner using LangGraph structured reasoning with real-time logging, and an AI verbal communication trainer using the Gemini API. I’ve completed 30+ AI/ML projects, published a research paper on Network Intrusion Detection (NCAET-24), and achieved top 5% in an AI certification program (Pro5.ai).