vik_bhati

Location: Blacksburg, VA, USA Remote: Yes, full-time, USA Willing to relocate: Yes Technologies: Python, Java, C++, SQL, PyTorch, TensorFlow, Hugging Face, LLM post-training (SFT, LoRA, GRPO), reinforcement learning, AI agent evaluation, RAG, GraphRAG, MCP, Spring Boot, FastAPI, Google Cloud, Docker, Kubernetes, TransformerLens, REST API Résumé/CV: https://vikrant-bhati.github.io/Resume/ Profile: https://vikrant-bhati.github.io/profile/ GitHub: https://github.com/vikrant-bhati/ LinkedIn: https://www.linkedin.com/in/vikrantbhati/ Email: [email protected]

I am an AI researcher at Virginia Tech with 7+ years of software engineering experience. I'm completing an M.S. in Computer Engineering, expected in December 2026, with a thesis on AI reliability and alignment.

My current research asks how AI agents reason when a task is underspecified and feedback is incomplete. I am leading TwoWorlds, an ongoing AI-for-science benchmark spanning geophysics, semiconductor diagnostics, thermal modeling, and battery monitoring. I also develop training pipelines for improving LLM reasoning on optimization problems.

Previously, at VIAVI Solutions, I implemented and deployed reinforcement learning to tune equalizer parameters, reducing bit error rate from 10^-7 to 10^-10 in 1.6 Tbps network-testing workflows. I also built a RAG application for querying proprietary technical data.

At Fiserv, I built a Google Cloud platform that ingested and validated 300M+ records weekly, developed Clover POS features, and created a database encryption framework adopted across 120+ applications.

I'm looking for full-time AI/ML engineering, research engineering, or software engineering roles in the US, particularly involving agent reliability, LLM evaluation and post-training, or production ML systems. I enjoy work that connects research experiments with dependable software.