Rajat Ghosh, Ph.D.

Highlights

• GenAI Platforms: Built company-scale LLM agents (code search, PR review, unit testing, support automation) adopted by hundreds of engineers/SREs.
• Business Impact: Delivered multi-million-dollar annual savings via productivity gains and vendor replacement.
• Post-Training: Led RLHF/post-training (GRPO/KTO/SFT), evaluation, and safety pipelines for production readiness.
• Cross-Functional: Led enterprise rollout with Product/Infra/Security/Legal; drove compliance and procurement readiness.
• Research: Publications, patents, and open-source contributions in LLM alignment, benchmarking, and AI safety.

Professional Summary

I am a senior technical leader with a sustained record of building company-scale AI platforms from early research through production adoption. My work spans LLM systems, agentic workflows, post-training and alignment, evaluation, and infrastructure optimization, with measurable business impact and durable organizational leverage.

• Platform scope: Delivered GenAI systems (developer productivity and support automation) adopted across teams and geographies with production metrics.
• Technical depth: Built end-to-end pipelines across RAG, agentic workflows, reasoning-oriented post-training (GRPO/KTO/SFT), evaluator models, red-teaming, and GPU-efficient serving.
• Leadership: Built and led multi-disciplinary teams; established execution rhythms (OKRs, stakeholder reviews, EA/GA rollouts) and operated effectively across Product, Infra, Security, Legal, and executive stakeholders.

Skills

• Domains: Large Language Models, ML Architecture, Reinforcement Learning, Retrieval Augmented Generation, API Design, LLMOps, Optimal Control
• ML Libraries: HuggingFace, Transformers, Gym, PyTorch, Jax
• Databases: SQL, Milvus, Elasticsearch
• Languages: C++, Python, Go, JavaScript, Bash, Java
• Technologies: PySpark, Docker, Prometheus, Jenkins, Postman
• Platforms: AWS, GCP, NVIDIA

Professional Experience

Technical Director, AI – Nutanix (September 2025 – Present)

As Technical Director, I lead the post-training and evaluation product portfolio for the Nutanix GenAI platform, driving model quality, reliability, and readiness for production.

• Lead a team of 7 engineers to deliver post-training and evaluation products for the Nutanix GenAI platform (quality, reliability, production readiness).
• Owned evaluation strategy across offline + human feedback loops; standardized release gates, regression tracking, and model readiness criteria.
• Built AI-driven data capabilities including dataset curation, synthetic data generation, and labeling to accelerate fine-tuning and alignment workflows.

Staff Scientist – Nutanix (May 2021 – August 2025)

As Tech Lead Data Scientist, I spearheaded the design and delivery of production-ready generative AI agents for customer service and developer productivity. I defined the architectures, built initial implementations, and established rigorous evaluation pipelines to validate quality and reliability.

• Architected internal coding assistant agents utilizing open-source LLMs, enhancing productivity by 20% ($100M annual impact).
• Developed RAG-based support assistants using fine-tuned LLMs ($20M annual impact).
• Engineered reinforcement learning strategies for distributed infrastructure management.
• Created explainability metrics and object detection inference APIs using YOLO v5.

Co-Founder & CEO – AdeptDC, Inc. (July 2015 – April 2021)

• Founded a startup with a vision to optimize data center energy efficiency through RL-based dynamic control.
• Managed a 7-member team, securing ~$1M funding from VC and NSF.

Education

• Ph.D., Georgia Institute of Technology, 2013 (Dissertation)
• M.S., Georgia Institute of Technology, 2010
• B.S., Indian Institute of Technology, Kharagpur, 2008

Selected Publications on Reinforcement Learning

Action Shapley: A Training Data Selection Metric for World Model in Reinforcement Learning

Rajat Ghosh, Debojyoti Dutta

arXiv · 2026

Paper

Selected Publications on LLMs

Go-UT-Bench: A Fine-Tuning Dataset for LLM-Based Unit Test Generation in Go

Yashshi Pipalani, Hritik Raj, Rajat Ghosh, Vaishnavi Bhargava, Debojyoti Dutta

arXiv · 2025

Paper

A Multi-Agent Framework for Stateful Inference-Time Search

Arshika Lalan, Rajat Ghosh, Aditya Kolsur, Debojyoti Dutta

arXiv · 2025

Paper

RANGER: Repository-Level Agent for Graph-Enhanced Retrieval

Pratik Shah, Rajat Ghosh, Aryan Singhal, Debojyoti Dutta

arXiv · 2025

Paper

BAR Conjecture: the Feasibility of Inference Budget-Constrained LLM Services with Authenticity and Reasoning

Jinan Zhou, Rajat Ghosh, Vaishnavi Bhargava, Debojyoti Dutta, Aryan Singhal

arXiv · 2025

Paper

Predictive Scaling Laws for Efficient GRPO Training of Large Reasoning Models

Datta Nimmaturi, Vaishnavi Bhargava, Rajat Ghosh, Debojyoti Dutta

arXiv · 2025

Paper

Code2json: Can a Zero-Shot LLM Extract Code Features for Code RAG?

Aryan Singhal, Rajat Ghosh, Ria Mundra, Harshil Dadlani, Debojyoti Dutta

ICLR Deep Learning for Code Workshop · 2025

Paper

CPP-UT-Bench: Complex Unit Tests in C++ via LLMs

Vaishnavi Bhargava, Rajat Ghosh, Debojyoti Dutta

NeurIPS Workshop on Statistical Foundations of LLMs and Foundation Models · 2024

Paper

Efficient Alignment of LLMs via Data Sampling

Amrit Khera, Rajat Ghosh, Debojyoti Dutta

NeurIPS ENLSP Workshop · 2024

Paper

Consortium Publications from MLCommons

AILuminate Security Introducing v0.5 of the Jailbreak Benchmark from MLCommons

Consortium author (49 authors). Contributed to the multi-modal dataset curation.

arXiv (MLCommons) · 2025

Paper

AILuminate: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons

Consortium author (82 authors). Contributed to the evaluation pipeline development.

arXiv (MLCommons) · 2025

Paper

Introducing v0.5 of the AI Safety Benchmark from MLCommons

Consortium author (100 authors). Contributed to the dataset curation.

arXiv (MLCommons) · 2024

Paper

Selected Publications from Doctoral Research

Proper Orthogonal Decomposition-Based Modeling Framework for Improving Spatial Resolution of Measured Temperature Data

Rajat Ghosh, Yogendra Joshi

IEEE CPMT · 2014

Paper

Rapid Temperature Predictions in Data Centers using Multi-Parameter Proper Orthogonal Decomposition

Rajat Ghosh, Yogendra Joshi

Numerical Heat Transfer, Part A · 2014

Paper

Error Estimation in POD-Based Dynamic Reduced-Order Thermal Modeling of Data Centers

Rajat Ghosh, Yogendra Joshi

International Journal of Heat and Mass Transfer · 2013

Paper

Effect of rack server population on temperatures in data centers

Rajat Ghosh, Yogendra Joshi

13th InterSociety Conference on Thermal and Thermomechanical Phenomena in Electronic Systems · 2013

Paper

Designing Oscillating Cilia That Capture or Release Microscopic Particles

Rajat Ghosh, Gavin A. Buxton, Berk Usta, Anna C. Balazs, Alexander Alexeev

Langmuir · 2010

Paper

Patents

• Generating rules for managing an infrastructure from natural-language expressions, U.S. Utility Patent Pending, 2023. (link)

• Anomaly detection with model hyperparameter selection, U.S. Utility Patent Pending, 2023. (link)

• Data-driven evaluation of training action space for reinforcement learning, U.S. Utility Patent Pending, 2023. (link)

• Systems and Methods for Intelligent Controls in Data Centers, U.S. Utility Patent, 2019. (link)

Honors and Awards

• NSF Small Business Innovation Research Award, 2015. (link)
• Georgia Research Alliance Phase IA Award, 2014.
• NSF I-Corps Award, 2014.

Invited Talks/Panels

• "Agentic AI: Future Challenges and Opportunities", AAAI Workshop, March 2025. (link)
• "AI Coding Best Practices", Nutanix Tech Talk, July 2025. (link)
• "CUDA Programming", Nutanix Tech Talk, June 2025. (link)
• "From Code to Reality: A Day in the Life of a Nutanix AI Superuser", Nutanix Annual Conference, May 2025. (link)
• "Deploying your First GenAI Workload with GPT-in-a-Box", Nutanix Annual Conference, May 2024. (link)
• "Nutanix SupportGPT", Nutanix Tech Talk, April 2024. (link)
• "LLM Codegen at Nutanix", Nutanix Tech Talk, March 2024. (link)

Technical Blogs

• "Run LLM Pre-Training on Nutanix", Nutanix Developer Blog, September 2023. (link)
• "LLM Model Selection and Inference Benchmarking on Nutanix Cloud Platform", Nutanix Developer Blog, August 2023. (link)
• "It's Time For AI: LLM-Based Unit Tests for OpenSource Repositories", Nutanix Tech Center Blog. (link)
• "It's Time For AI: Providing Delightful Support with AI at Nutanix", Nutanix Tech Center Blog. (link)
• "It's Time For AI: How Nutanix Implemented an LLM Agent", Nutanix Tech Center Blog. (link)
• "It's Time For AI: Benchmark a Coding Agent on Nutanix", Nutanix Tech Center Blog. (link)
• "RLHF on Nutanix Cloud Platform", Nutanix Developer Blog, November 2023. (link)
• "PEFT for LLM Fine-Tuning on Nutanix Cloud Platform", Nutanix Developer Blog, July 2023. (link)
• "Run Attention on Nutanix", Nutanix Developer Blog, May 2023. (link)
• "Train Transformer Models on Nutanix", Nutanix Developer Blog, May 2023. (link)

Teaching Experience

• Thermodynamics, Georgia Tech, 2013
• Experimental Methodology, Georgia Tech, 2012-2013
• Introduction to Energy Systems Engineering, Georgia Tech, 2012
• Creative Decisions and Design, Georgia Tech, 2009

Professional Services

• Reviewer for NeurIPS 2025