Akash Srivastava

I am the Director and Chief Architect of Core AI at IBM, where I lead AgentOps—a platform for building, evaluating, and optimizing agentic AI systems.

I am a Principal Investigator at the MIT-IBM Watson AI Lab, where I led the post-training team for Granite. I also founded the Red Hat AI Innovation Team (now in advisory role), where we created InstructLab—an open-source framework for LLM customization.

I obtained my PhD at the University of Edinburgh, where I worked with Prof Charles Sutton and Prof Michael U. Gutmann on variational inference and generative models.

Research

My research interests include agentic AI systems, inference-time scaling and reasoning, post-training and model customization, and generative modeling.

Selected Press

All media coverage

Publications

Inference-Time Scaling, Sampling, & Probabilistic Inference (LLMs)

Mitigating Premature Exploitation in Particle-based Monte Carlo for Inference-Time Scaling
Mitigating Premature Exploitation in Particle-based Monte Carlo for Inference-Time Scaling
Giorgio Giannone, Guangxuan Xu, Nikhil Shivakumar Nayak, Rohan Mahesh Awhad, Shivchander Sudalairaj, Kai Xu, Akash Srivastava
arXiv preprint, 2025
[paper]
Rollout Roulette: A Probabilistic Inference Approach to Inference-Time Scaling of LLMs
Rollout Roulette: A Probabilistic Inference Approach to Inference-Time Scaling of LLMs
Isha Puri, Shivchander Sudalairaj, Guangxuan Xu, Kai Xu, Akash Srivastava
arXiv preprint, 2025
[paper]
Inference-Time Scaling of Diffusion Language Models with Particle Gibbs Sampling
Inference-Time Scaling of Diffusion Language Models with Particle Gibbs Sampling
Meihua Dang, Jiaqi Han, Minkai Xu, Kai Xu, Akash Srivastava, Stefano Ermon
arXiv preprint, 2025
[paper]
Variational Russian Roulette for Deep Bayesian Nonparametrics
Variational Russian Roulette for Deep Bayesian Nonparametrics
Kai Xu, Akash Srivastava, Charles Sutton
ICML, 2019
[paper]

Alignment, Preference Learning, & Human Feedback

LAB: Large-Scale Alignment for ChatBots
LAB: Large-Scale Alignment for ChatBots
Shivchander Sudalairaj, Abhishek Bhandwaldar, Aldo Pareja, Kai Xu, David D. Cox, Akash Srivastava
arXiv preprint, 2024
[paper]
Dr. SoW: Density Ratio of Strong-over-Weak LLMs for Reducing the Cost of Human Annotation
Dr. SoW: Density Ratio of Strong-over-Weak LLMs for Reducing the Cost of Human Annotation
Guangxuan Xu, Kai Xu, Shivchander Sudalairaj, Hao Wang, Akash Srivastava
arXiv preprint, 2024
[paper]
Value-Augmented Sampling for Language Model Alignment and Personalization
Value-Augmented Sampling for Language Model Alignment and Personalization
Seungwook Han, Idan Shenfeld, Akash Srivastava, Yoon Kim, Pulkit Agrawal
arXiv preprint, 2024
[paper]
Post-Processing Private Synthetic Data for Improving Utility on Selected Measures
Post-Processing Private Synthetic Data for Improving Utility on Selected Measures
Hao Wang, Shivchander Sudalairaj, John Henning, Kristjan Greenewald, Akash Srivastava
NeurIPS, 2023
[paper]

Model Merging, Fine-Tuning, & Parameter Efficiency

Activation-Informed Merging of Large Language Models
Activation-Informed Merging of Large Language Models
Amin Heyrani Nobari, Kaveh Alim, Ali ArjomandBigdeli, Akash Srivastava, Faez Ahmed, Navid Azizan
arXiv preprint, 2025
[paper]
Sculpting Subspaces: Constrained Full Fine-Tuning in LLMs for Continual Learning
Sculpting Subspaces: Constrained Full Fine-Tuning in LLMs for Continual Learning
Nikhil Shivakumar Nayak, Krishnateja Killamsetty, Ligong Han, Abhishek Bhandwaldar, Prateek Chanda, Kai Xu, Hao Wang, Aldo Pareja, Oleg Silkin, Mustafa Eyceoz, Akash Srivastava, et al.
arXiv preprint, 2025
[paper]
Squat: Subspace-Orthogonal KV Cache Quantization
Squat: Subspace-Orthogonal KV Cache Quantization
Hao Wang, Ligong Han, Kai Xu, Akash Srivastava
arXiv preprint, 2025
[paper]
Hopscotch: Discovering and Skipping Redundancies in Language Models
Hopscotch: Discovering and Skipping Redundancies in Language Models
Mustafa Eyceoz, Nikhil Shivakumar Nayak, Hao Wang, Ligong Han, Akash Srivastava
arXiv preprint, 2025
[paper]
Unveiling the Secret Recipe: A Guide for Supervised Fine-Tuning Small LLMs
Unveiling the Secret Recipe: A Guide for Supervised Fine-Tuning Small LLMs
Aldo Pareja, Nikhil Shivakumar Nayak, Hao Wang, Krishnateja Killamsetty, Shivchander Sudalairaj, Wenlong Zhao, Seungwook Han, Abhishek Bhandwaldar, Guangxuan Xu, Kai Xu, Akash Srivastava, et al.
arXiv preprint, 2024
[paper]

Diffusion Models & Generative Modeling

SODA: Spectral Orthogonal Decomposition Adaptation for Diffusion Models
SODA: Spectral Orthogonal Decomposition Adaptation for Diffusion Models
Xinxi Zhang, Song Wen, Ligong Han, Felix Juefei-Xu, Akash Srivastava, Junzhou Huang, Vladimir Pavlovic, Hao Wang, Molei Tao, Dimitris Metaxas
WACV, 2025
[paper]
DICE: Discrete Inversion Enabling Controllable Editing for Multinomial Diffusion and Masked Generative Models
DICE: Discrete Inversion Enabling Controllable Editing for Multinomial Diffusion and Masked Generative Models
Xiaoxiao He, Quan Dao, Ligong Han, Song Wen, Minhao Bai, Di Liu, Han Zhang, Martin Renqiang Min, Felix Juefei-Xu, Chaowei Tan, Akash Srivastava, et al.
arXiv preprint, 2024
[paper]
Aligning Optimization Trajectories with Diffusion Models for Constrained Design Generation
Aligning Optimization Trajectories with Diffusion Models for Constrained Design Generation
Giorgio Giannone, Akash Srivastava, Ole Winther, Faez Ahmed
NeurIPS, 2023
[paper]

Synthetic Data, Privacy, & Density Ratio Estimation

Differentially Private Synthetic Data Generation for Relational Databases
Differentially Private Synthetic Data Generation for Relational Databases
Kaveh Alimohammadi, Hao Wang, Ojas Gulati, Akash Srivastava, Navid Azizan
arXiv preprint, 2024
[paper]
Private Synthetic Data Meets Ensemble Learning
Private Synthetic Data Meets Ensemble Learning
Haoyuan Sun, Navid Azizan, Akash Srivastava, Hao Wang
arXiv preprint, 2023
[paper]
Estimating the Density Ratio Between Distributions with High Discrepancy
Estimating the Density Ratio Between Distributions with High Discrepancy
Akash Srivastava, Seungwook Han, Kai Xu, Benjamin Rhodes, Michael U. Gutmann
arXiv preprint, 2023
[paper]
Generative Ratio Matching Networks
Generative Ratio Matching Networks
Akash Srivastava, Michael U. Gutmann, Kai Xu, Charles Sutton
ICLR, 2020
[paper]

Red-Teaming, Robustness, & Evaluation

Curiosity-Driven Red-Teaming for Large Language Models
Curiosity-Driven Red-Teaming for Large Language Models
Zhang-Wei Hong, Idan Shenfeld, Tsun-Hsuan Wang, Yung-Sung Chuang, Aldo Pareja, James Glass, Akash Srivastava, Pulkit Agrawal
arXiv preprint, 2024
[paper]
On the Importance of Calibration in Semi-Supervised Learning
On the Importance of Calibration in Semi-Supervised Learning
Charlotte Loh, Rumen Dangovski, Shivchander Sudalairaj, Seungwook Han, Ligong Han, Leonid Karlinsky, Marin Soljacic, Akash Srivastava
arXiv preprint, 2022
[paper]
Mitigating Confirmation Bias in Semi-Supervised Learning via Efficient Bayesian Model Averaging
Mitigating Confirmation Bias in Semi-Supervised Learning via Efficient Bayesian Model Averaging
Charlotte Loh, Rumen Dangovski, Shivchander Sudalairaj, Seungwook Han, Ligong Han, Leonid Karlinsky, Marin Soljacic, Akash Srivastava
TMLR, 2023
[paper]

Foundation Models for Planning, Design, & Systems

Compositional Foundation Models for Hierarchical Planning
Compositional Foundation Models for Hierarchical Planning
Anurag Ajay, Seungwook Han, Yilun Du, Shuang Li, Abhi Gupta, Tommi Jaakkola, Josh Tenenbaum, Leslie Kaelbling, Akash Srivastava, Pulkit Agrawal
NeurIPS, 2023
[paper] [project]
Learning Joint Representations of Design and Performance Spaces
Learning Joint Representations of Design and Performance Spaces
Amin Heyrani Nobari, Akash Srivastava, Dan Gutfreund, Kai Xu, Faez Ahmed
arXiv preprint, 2024
[paper]
Beyond Statistical Similarity: Rethinking Metrics for Deep Generative Models in Engineering Design
Beyond Statistical Similarity: Rethinking Metrics for Deep Generative Models in Engineering Design
Lyle Regenwetter, Akash Srivastava, Dan Gutfreund, Faez Ahmed
Computer-Aided Design, 2023
[paper]

Lifelong Learning, Program Synthesis, & Modularity

Houdini: Lifelong Learning as Program Synthesis
Houdini: Lifelong Learning as Program Synthesis
Lazar Valkov, Dipak Chaudhari, Akash Srivastava, Charles Sutton, Swarat Chaudhuri
NeurIPS, 2018
[paper]
A Probabilistic Framework for Modular Continual Learning
A Probabilistic Framework for Modular Continual Learning
Lazar Valkov, Akash Srivastava, Swarat Chaudhuri, Charles Sutton
arXiv preprint, 2023
[paper]
Synthesis of Differentiable Functional Programs for Lifelong Learning
Synthesis of Differentiable Functional Programs for Lifelong Learning
Lazar Valkov, Dipak Chaudhari, Akash Srivastava, Charles Sutton, Swarat Chaudhuri
arXiv preprint, 2018
[paper]

Core Generative Modeling & Bayesian ML

VEEGAN: Reducing Mode Collapse in GANs using Implicit Variational Learning
VEEGAN: Reducing Mode Collapse in GANs using Implicit Variational Learning
Akash Srivastava, Lazar Valkov, Chris Russell, Michael U. Gutmann, Charles Sutton
NeurIPS, 2017
[paper] [code]
Autoencoding Variational Inference For Topic Models
Autoencoding Variational Inference For Topic Models
Akash Srivastava, Charles Sutton
ICLR, 2017
[paper] [code]
Fast and Scalable Bayesian Deep Learning by Weight-Perturbation in Adam
Fast and Scalable Bayesian Deep Learning by Weight-Perturbation in Adam
Mohammad Emtiyaz Khan, Didrik Nielsen, Voot Tangkaratt, Wu Lin, Yarin Gal, Akash Srivastava
ICML, 2018
[paper]
Deep Generative Modelling for Amortised Variational Inference
Akash Srivastava
PhD Thesis, University of Edinburgh, 2019
[paper]

Applied / Cross-Domain Modeling

Learning to Deliver: A Foundation Model for Vehicle Routing
Learning to Deliver: A Foundation Model for Vehicle Routing
Samuel JK Chin, Matthias Winkenbach, Akash Srivastava
arXiv preprint, 2024
[paper]
Urban Context and Delivery Performance
Urban Context and Delivery Performance
Maxwell Schrader, Navish Kumar, Esben Sorig, Soonmyeong Yoon, Akash Srivastava, Kai Xu, Maria Astefanoaei, Nicolas Collignon
arXiv preprint, 2024
[paper]

For complete citation information and PDFs, visit my Google Scholar profile.