Vision
My goal is to make LLM agents trustworthy in the loop: they should externalize their reasoning into tools (code, tests, search, structured checks), verify intermediate results, and recover when things go wrong.
Core themes
1) Verification-aware prompting and pipelines
- Design generator–validator workflows that detect errors early and enforce explicit checks.
- Use unit tests, static checks, and consistency constraints to reduce silent failures.
2) RL for tool-use policies
- Learn when to call tools, how to compose tool outputs, and when to backtrack.
- Optimize for reliability metrics (accuracy under distribution shift, robustness, and calibration).
3) Neuro-symbolic representations for dialog + code
- Represent dialog and code with operational/axiomatic semantics to enable structured validation.
- Combine symbolic constraints with neural generation to improve interpretability and correctness.
Selected work & experience
Graduate Research Assistant — Stony Brook University (01/2023–05/2027)
- Built LLM-based validation agents to detect errors in open-ended collaborative coding tasks.
- Implemented verification-aware generator–validator pipelines for static and real-time evaluation.
- Developed tool-augmented pipelines using knowledge graphs and multimodal NLP.
Applied Scientist Intern — Amazon AWS (09/2025–12/2025)
- Empirically evaluated the “GEPA” Automatic Prompt Optimization (APO) framework.
- Benchmarked across 10 production LLMs (1B–405B), validating structured reasoning improvements up to 12%.
- Analyzed cross-model prompt transfer (mid-sized → larger models) for reusable optimization.
ML Research Intern — Nokia Bell Labs (06/2025–08/2025)
- Developed a multi-agent LLM pipeline using OPC UA to extract insights from industrial telemetry.
- Integrated knowledge graphs + function calling to correlate events across system layers.
Publications
2025
- Shaik, H., Villuri, G., & Doboli, A. (2025). Two Large Language Model-based Methods to Validate Open-Ended Problem Solving in Teams. (AIMS 2025)
- Shaik, H., Villuri, G., & Doboli, A. (2025). An Overview of LLMs and a Novel, LLM-Based Cognitive Architecture for Solving Open-Ended Problems. (PDF)
- Shaik, H., Villuri, G., & Doboli, A. (2025). Concept Combinations with Generator and Validator Agents Prompted Using Insights from Concept Networks. (Conference)
- Villuri, G., & Doboli, A. (2025). An Experimental Study on the Interpretability of Transformer Models for Dialog Understanding. (IEEE)
- Villuri, G., Shaik, H., & Doboli, A. (2025). A Stacked Multi-Layered Perceptron - LLM Model for Extracting the Relations in Textual Descriptions. (IEEE)
- Villuri, G., Doboli, A., & Pallapu, H. R. (2025). Towards Semantic Classification: An Experimental Study on Automated Understanding of the Meaning of Verbal Utterances. (IEEE Annual Computing and Communication Workshop and Conference, 2025)
2024
- Villuri, G., & Doboli, A. (2024). An Overview and Discussion of the Suitability of Existing Speech Datasets to Train Machine Learning Models for Collective Problem Solving. (arXiv:2412.18489)
- Villuri, G., & Doboli, A. (2024). Using Speech Data to Automatically Characterize Team Effectiveness to Optimize Power Distribution in Internet-of-Things Applications. (IEEE CITDS 2024)
- Pallapu, H. R., Villuri, G., Doboli, A., & Doboli, S. (2024). Automatically Understanding Human Behavior for IoT Applications with Optimized Human-in-the-Loop Control.
- Villuri, G., & Doboli, A. (2024). Towards Semantic Classification of Dialog using Contextual Prediction Networks. (Scholar)
Contact
Email: villurignanesh@gmail.com | Google Scholar | LinkedIn