Hemanth Vankudoth

I’m an undergraduate at the Indian Institute of Technology (BHU), studying B.Tech in Mechanical Engineering, and I focus my work on AI systems that actually ship. From “AI for Programmers” and soft-computing coursework to hands-on GenAI projects, I enjoy turning models into reliable products.

In my AI—Software Engineer Intern role, I automated end-to-end integration tests and introduced quality gates in GitHub Actions CI/CD, cutting post-release bugs by 40% and accelerating development velocity by 50%. I also built a real-time notification API and enhanced the Upcaster microservice (Voyage AI and Vertex AI) to handle edge cases, improving reliability and accuracy of automated test case generation.

My projects push low-latency, scalable architectures: ScamShield uses Gemini 2.0 Flash with a Hybrid RAG (Qdrant) to detect multimodal financial fraud, while FinGPT fine-tunes TinyLlama with LoRA/QLoRA (PEFT/TRL) and deploys a low-latency inference stack with llama.cpp, FastAPI, and React. I also build end-to-end recommendation and ML pipelines using Docker, CI/CD, DVC, AWS, and reproducible tooling—always with an emphasis on reliability.