Bridging AI research and production systems

I build and optimize AI systems that move from research to production.

I am a Technical Lead at RELAI and a PhD graduate from the University of Maryland. I focus on agentic platforms, LLM evaluation, system optimization, and reliability infrastructure for real-world AI products.

My work combines deep research grounding (ICLR, ICML, NeurIPS publications) with hands-on leadership building production-ready systems.

View selected work Reach out

About

I specialize in building AI systems where quality improves continuously—from evaluation and robustness testing to optimization, deployment, and operational monitoring.

My core strength is translating advanced research into practical engineering systems that teams can ship, maintain, and scale with confidence.

I'm particularly interested in LLM/agent evaluation frameworks, production architecture for AI products, and technical leadership at the intersection of research and execution.

Selected Work

Agentic Platform for Production AI Workflows

Problem: Teams needed a reliable way to move agent experiments into production without brittle glue code.

Built: Led architecture for a reusable orchestration layer with observability, guardrails, agent optimization loops, and deployment automation.

Why it matters: Reduced iteration friction for product teams and created a common backbone for shipping robust AI features faster.

LLM Evaluation & Reliability Pipelines

Problem: Model quality was hard to compare consistently across prompts, tasks, and releases.

Built: Designed automated evaluation pipelines for regression checks, failure analysis, benchmark tracking, and iterative agent/system tuning.

Why it matters: Improved release confidence and made reliability measurable during development instead of after incidents.

Research-to-Production AI Systems

Problem: Cutting-edge research often stalls before it reaches users.

Built: Bridged applied research and engineering by converting reliability methods into deployable systems and practical workflows.

Why it matters: Enabled faster translation of research ideas into product capabilities with stronger safety and performance characteristics.

Timeline

2024 - Present
Technical Lead at RELAI
- Leading design and delivery of agentic AI platforms focused on evaluation, robustness, and deployment readiness.
- Driving engineering execution across reliability infrastructure, multi-agent workflows, optimization pipelines, and production system architecture.
2020 - 2024
PhD in Electrical and Computer Engineering, University of Maryland, College Park
- Worked on reliability of machine learning systems and generative models with a focus on robust behavior under distribution shift.
2018 - 2020
M.Tech in Electrical Engineering, IIT Madras
- Master's thesis on image denoising with A. N. Rajagopalan. Completed a summer internship at KLA focused on super-resolution of SEM images.
2016 - 2018
Co-Founder, Tessact
- Built software for scalable frame-level video annotation to support contextual advertising and media compliance workflows.
2012 - 2016
B.Tech in Electrical Engineering, IIT Bombay
- First exposure to computer vision through courses taught by Ajit Rajwade.

Publications

Rethinking Artistic Copyright Infringements in the Era of Text-to-Image Generative Models

Mazda Moayeri, Samyadeep Basu, Sriram Balasubramanian, Priyatham Kattakinda, Atoosa Chegini, Soheil Feizi

International Conference on Learning Representations (ICLR), 2025

Paper

On mechanistic knowledge localization in text-to-image generative models

Samyadeep Basu, Keivan Rezaei, Priyatham Kattakinda, Vlad I. Morariu, Niranjan Zhao, Ryan A. Rossi, Soheil Feizi

Forty-first International Conference on Machine Learning (ICML), 2024

Paper

Fast Adversarial Attacks on Language Models In One GPU Minute

Vinu Sankar Sadasivan, Shoumik Saha, Gaurang Sriramanan, Priyatham Kattakinda, Atoosa Chegini, Soheil Feizi

International Conference on Learning Representations (ICLR), 2024

Paper

LLM-Check: Investigating Detection of Hallucinations in Large Language Models

Gaurang Sriramanan, Siddhant Bharti, Vinu Sankar Sadasivan, Shoumik Saha, Priyatham Kattakinda, Soheil Feizi

Advances in Neural Information Processing Systems (NeurIPS), 2024

Paper

FOCUS: Familiar Objects in Common and Uncommon Settings

Priyatham Kattakinda, Soheil Feizi

Proceedings of the 39th International Conference on Machine Learning, 2022

Paper Code Dataset

Unpaired Image Denoising

Priyatham Kattakinda, A. N. Rajagopalan

IEEE International Conference on Image Processing (ICIP), UAE (Virtual), October 2020

Paper Code Presentation Project Page

Preprints

Understanding and Mitigating Compositional Issues in Text-to-Image Generative Models

Arman Zarei, Keivan Rezaei, Samyadeep Basu, Mehrdad Saberi, Mazda Moayeri, Priyatham Kattakinda, Soheil Feizi

Paper

Understanding the Effect of using Semantically Meaningful Tokens for Visual Representation Learning

Neha Kalibhat, Priyatham Kattakinda, Arman Zarei, Nikita V Seleznev, Samuel Sharpe, Sanjiv Kumar, Soheil Feizi

Paper

Invariant Learning via Diffusion Dreamed Distribution Shifts

Priyatham Kattakinda, Alexander Levine, Soheil Feizi

Paper

Accomplishments

Outstanding Graduate Student Award [December 2023]
Recognized among the top 2% of graduate assistants at the University of Maryland in a given year.
Siemens Prize [September 2020]
Awarded for best academic record in Electrical Engineering at IIT Madras.
M K Achuthan Prize [April 2020]
Awarded for best first-year academic record in Electrical Engineering at IIT Madras.
All India Rank 80 in IIT-JEE [June 2012]
Secured rank 80 among approximately 470,000 candidates in IIT-JEE 2012.