Research

Research Vision

I work on the foundations and applications of compound AI and multi-agent systems — LLM-driven pipelines whose components retrieve, reason, plan, and act in coordination. At Amazon Stores Science, this translates to multi-agent customer-behavior models that jointly optimize search, selection, and econometrics. Three threads run through my research:

  1. Optimization & reward design for compound AI. How do we train and align multi-component LLM systems when end-to-end signals are sparse and credit assignment is hard? Recent work (Optimas, ICLR 2026; AgentDR, WWW 2026; RecLAIF) introduces globally-aligned local rewards, LLM-agent recommendation with implicit relations, and reinforcement learning from AI feedback for agentic pipelines.
  2. Reasoning over structured knowledge with LLMs. Bridging neural-symbolic reasoning and foundation models for retrieval, ranking, and complex logical queries (WWW 2024, ArXiv 2023, GT2Vec).
  3. Geometric representation learning at scale. Hyperbolic and non-Euclidean networks for hierarchical data, deployed in production search (NeurIPS 2023, WSDM 2022, NeurIPS 2021).

My work has shipped to Amazon Stores, appeared at ICLR / NeurIPS / KDD / WWW / WSDM / IJCAI, and is reflected in one US patent and several open-source releases under amazon-science.

Tutorials

Hyperbolic Networks: Theory, Architectures And Applications
Tutorial @ The Web Conference (WWW 2022)
Speakers: Nurendra Choudhary, Nikhil Rao, Karthik Subbian, Srinivasan H. Sengamedu, Chandan K. Reddy
[Tutorial Link]    [Teaser Video]

Tutorial @ 28th SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 2022)
Speakers: Nurendra Choudhary, Nikhil Rao, Karthik Subbian, Srinivasan H. Sengamedu, Chandan K. Reddy
[Tutorial Link]    [Teaser Video]

Tutorial @ 37th AAAI Conference on Artificial Intelligence (AAAI 2023)
Speakers: Nurendra Choudhary, Karthik Subbian, Srinivasan H. Sengamedu, Chandan K. Reddy
[Tutorial Link]    [Teaser Video]
Invited Talks
Invited talks on hyperbolic representation learning, knowledge-graph reasoning, and LLM/graph integration for search and recommendation at academic and industry venues (2023–present).

Patents

  • Index searching with non-Euclidean representations. S. Katariya, N. S. Rao, C. K. Reddy, K. Subbian, N. Choudhary.
    US Patent 11,810,174, 2023. [pdf]

Open Source

  • HGRAM — Hyperbolic graph meta-learning framework (NeurIPS 2023).

  • graph-lm-ensemble — Interpretable graph + language model ensemble for e-commerce search (WWW 2024).

  • hyperbolic-embeddings — Hyperbolic entity embeddings for product search (WSDM 2022, WWW 2021).

  • structure-aware-language-models — Graph-based multilingual LM for search relevance (KDD 2022).

  • LLM-KG-Reasoning — Complex logical reasoning over KGs with LLMs.

  • PERM-GaussianKG — Probabilistic entity reasoning over knowledge graphs (NeurIPS 2021).

  • GraphZoo — Toolkit for hyperbolic graph neural networks (WWW 2022).

Mentorship & Collaboration

I have collaborated with and mentored interns and visiting students at Amazon, and co-advised research projects with Ph.D. students at Virginia Tech, the University of Michigan, Stanford, and the National University of Singapore, leading to publications at ICLR, NeurIPS, KDD, WWW, IJCAI, and LoG. I am always happy to discuss collaborations on multi-agent systems, LLM optimization, and graph-grounded reasoning.

Services

Program committee / Reviewer
ICLR 2026, NeurIPS 2025, KDD 2025, WWW 2025, AAAI 2025, NeurIPS 2024, KDD 2024, WWW 2024, AAAI 2024, NeurIPS 2023, KDD 2023, WWW 2023, AAAI 2023, KDD 2022, AAAI 2021, NAACL 2021, ICDM 2020, EMNLP 2020, ACL 2020, ECIR 2019
Reviewer to Journals
IEEE Transactions on Knowledge and Data Engineering (TKDE)
ACM Transactions on Knowledge Discovery from Data (TKDD)
ACM Transactions on Intelligent Systems and Technology (TIST)
Geoscience Frontiers

Research Assistantships

  • Sanghani Center for Artificial Intelligence and Data Analytics, Virginia Tech

    [2019→2022]


  • Language Technologies Research Center, International Institute of Information Technology, Hyderabad

    [2016→2018]


  • Geotechnical Engineering Lab, International Institute of Information Technology, Hyderabad

    [2015]


  • Virtual Labs Engineering and Architecture Division, International Institute of Information Technology, Hyderabad

    [2014]