Shubham Agarwal

Portrait of Shubham Agarwal

Hi! I am a Ph.D. candidate in Computer Science at UC Berkeley, in the Sky Computing Lab and the EPIC Data Lab. I am advised by Prof. Ion Stoica and Prof. Aditya Parameswaran. I am also a Student Researcher at Google Cloud AI Research, co-hosted by Chun-Liang Li, Rui Meng and Violet Peng.

My research is on self-improving AI and systems. I build agents that continuously optimize systems such as LLM serving engines and GPU kernels, and agents that build new systems backed by tests and machine-checked proofs. I am a core developer and maintainer of SkyDiscover, our open-source framework for self-improving systems, which has 800+ GitHub stars and is used at Google, Uber and AMD. Within it, I lead DuoDiscover, now integrated into Google Antigravity.

Before Berkeley, I was a Research Associate at Adobe Research, working with Subrata Mitra and Shiv Kumar Saini on generative-AI serving and cloud reliability systems that shipped in three Adobe products. I received my Bachelor's in Computer Science from BITS Pilani in 2022.

The best way to reach me is email: shubham3@berkeley.edu.

More about my research

Continuous optimization

Agents that keep optimizing a running system, reaching 2.17× over AMD's expert-tuned SGLang.

1.92× Integrated into Google Antigravity faster vLLM serving

Agents write fast evaluators for slow benchmarks, then recheck on the real system. First on 121 of 235 SOL-Exec GPU kernels.

96% EvoX at COLM 2026 of 196 tasks

AdaEvolve tunes the search as it runs; EvoX lets the LLM rewrite it. Together, they beat every open framework.

Building trustworthy systems

Agents that build new systems backed by tests and machine-checked proofs.

7 of 7 NeurIPS 2026 Oral specs proved

Agents write code with its Rocq proof, for distributed key-value stores. The code runs up to 3× faster than hand-verified code.

up to 2.7× NeurIPS 2026 Workshop vs. prior Mac LLM servers

Agents build a system per workload and turn each caught shortcut into a test. Key-value stores beat FASTER on 14 of 18 workloads.

News

Nov 2026 Advanced to Ph.D. candidacy at the beginning of my second year.
Nov 2026 Talks on SkyDiscover at the SGLang Summit, Microsoft Azure Research, AMD and Inferact.
Oct 2026 DuoDiscover is out on arXiv.
Sep 2026 Inductive Deductive Synthesis accepted as an Oral at NeurIPS 2026; RCStat also accepted.
Aug 2026 Arming Data Agents with Tribal Knowledge accepted at VLDB 2026.
Jul 2026 Best Paper Award at the ICML 2026 FAGEN Workshop for Fantastic Adaptive Taxonomies.
Jul 2026 EvoX accepted at COLM 2026.
Jun 2026 Joined Google Cloud AI Research as a Student Researcher.
Apr 2026 SkyDiscover accepted as a Spotlight demo at ACM CAIS 2026, where I co-organized the AI Agents for Discovery in the Wild workshop.
Mar 2026 Released SkyDiscover, our open-source framework for self-improving systems.
Aug 2025 Started my Ph.D. at UC Berkeley.
Jul 2022 Joined Adobe Research as a Research Associate.

Publications 17

* equal contribution. Full list on Google Scholar.

  1. NeurIPS'26
    Inductive Deductive Synthesis: Co-Evolving Code-and-Proof Generation
    Shubham Agarwal*, Alexander Krentsel*, Shu Liu*, Mert Cemri*, Audrey Cheng, Rui Meng, et al.
    The 40th Conference on Neural Information Processing Systems (NeurIPS) 2026 Oral
  2. NeurIPS'26
    RCStat: A Statistical Framework of Relative Contextualization in Transformers
    Debabrata Mahapatra, Shubham Agarwal, Apoorv Saxena, Subrata Mitra
    The 40th Conference on Neural Information Processing Systems (NeurIPS) 2026
  3. COLM'26
    EvoX: Meta-Evolution for Automated Discovery
    Shu Liu*, Shubham Agarwal*, Monishwaran Maheswaran, Mert Cemri, Qiuyang Mang, et al.
    The 3rd Conference on Language Modeling (COLM) 2026
  4. VLDB'26
    Arming Data Agents with Tribal Knowledge
    Shubham Agarwal*, Asim Biswal*, Sepanta Zeighami*, Alvin Cheung, Joseph E. Gonzalez, et al.
    Proceedings of the VLDB Endowment (VLDB) 2026
  5. CAIS'26
    SkyDiscover: A Flexible, Adaptive Framework for AI-Driven Scientific and Algorithmic Discovery
    Shu Liu*, Mert Cemri*, Shubham Agarwal*, Alexander Krentsel, Ashwin Naren, Qiuyang Mang, et al.
    The 1st ACM Conference on AI and Agentic Systems (CAIS) 2026, Demo Spotlight
  6. ICLR'26
    FlowCast: Trajectory Forecasting for Scalable Zero-Cost Speculative Flow Matching
    Divya Jyoti Bajpai, Shubham Agarwal*, Apoorv Saxena*, Kuldeep Sharad Kulkarni, et al.
    The 14th International Conference on Learning Representations (ICLR) 2026
  7. CIDR'26
    Supporting Our AI Overlords: Redesigning Data Systems to be Agent-First
    Shu Liu, Soujanya Ponnapalli, Shreya Shankar, …, Shubham Agarwal, et al.
    The 16th Conference on Innovative Data Systems Research (CIDR) 2026
  8. Middleware'25
    Argus: Quality-Aware High-Throughput Text-to-Image Inference Serving System
    Shubham Agarwal, Subrata Mitra, Saud Iqbal
    The 26th ACM/IFIP International Middleware Conference (Middleware) 2025
  9. SIGMOD'25
    Cache-Craft: Managing Chunk-Caches for Efficient Retrieval-Augmented Generation
    Shubham Agarwal*, Sai Sundaresan*, Subrata Mitra, Debabrata Mahapatra, Archit Gupta, et al.
    International Conference on Management of Data (SIGMOD) 2025
  10. UAI'25
    Root Cause Analysis of Failures from Partial Causal Structures
    Azam Ikram*, Kenneth Lee*, Shubham Agarwal, Shiv Kumar Saini, Saurabh Bagchi, Murat Kocaoglu
    The 41st Conference on Uncertainty in Artificial Intelligence (UAI) 2025
  11. ECCV'24
    ReCon: Training-Free Acceleration for Text-to-Image Synthesis with Retrieval of Concept Prompt Trajectories
    Chen-Yi Lu*, Shubham Agarwal*, Mehrab Tanjim, Kanak Mahadik, Anup Rao, Subrata Mitra, et al.
    The 18th European Conference on Computer Vision (ECCV) 2024
  12. PAKDD'24
    ScaleViz: Scaling Visualization Recommendation Models on Large Data
    Ghazi Shazan Ahmad, Shubham Agarwal, Subrata Mitra, Ryan Rossi, Manav Doshi, et al.
    The 28th Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD) 2024 Oral
  13. NSDI'24
    Approximate Caching for Efficiently Serving Text-to-Image Diffusion Models
    Shubham Agarwal, Subrata Mitra, Sarthak Chakraborty, Srikrishna Karanam, et al.
    The 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI) 2024
  14. FSE'23
    Outage-Watch: Early Prediction of Outages using Extreme Event Regularizer
    Shubham Agarwal, Sarthak Chakraborty, Shaddy Garg, Sumit Bisht, Chahat Jain, et al.
    The 31st ACM Symposium on the Foundations of Software Engineering (ESEC/FSE) 2023
  15. ASE'23
    ESRO: Experience Assisted Service Reliability against Outages
    Sarthak Chakraborty, Shubham Agarwal, Shaddy Garg, Abhimanyu Sethia, et al.
    The 38th IEEE/ACM International Conference on Automated Software Engineering (ASE) 2023
  16. WWW'23
    CausIL: Causal Graph for Instance Level Microservice Data
    Sarthak Chakraborty, Shaddy Garg*, Shubham Agarwal*, Ayush Chauhan, Shiv Saini
    The ACM Web Conference (WWW) 2023
  17. SIGMOD'23
    Fast Natural Language-Based Data Exploration with Samples
    Shubham Agarwal, Gromit Yeuk-Yin Chan, Shaddy Garg, Tong Yu, Subrata Mitra
    International Conference on Management of Data (SIGMOD) 2023, Demo

Preprints 7

  1. arXiv'26
    DuoDiscover: Scaling Autonomous Discovery Beyond the Evaluation Bottleneck
    Shubham Agarwal*, Akshat Gupta*, Mike Heddes*, Dilara Soylu*, Rui Meng, et al.
    arXiv preprint 2026
  2. arXiv'26
    AdaEvolve: Adaptive LLM Driven Zeroth-Order Optimization
    Mert Cemri*, Shubham Agarwal*, Akshat Gupta, Shu Liu, Audrey Cheng, Qiuyang Mang, et al.
    arXiv preprint 2026
  3. arXiv'26
    The Time is Here for Just-in-Time Systems: Challenges and Opportunities
    Shu Liu*, Alexander Krentsel*, Shubham Agarwal*, Mert Cemri*, Ziming Mao, et al.
    arXiv preprint 2026; NeurIPS 2026 ML for Systems Workshop
  4. arXiv'26
    Fantastic Adaptive Taxonomies and How to Use Them
    Mert Cemri*, Andrei Cojocaru*, Melissa Pan, Shu Liu, Shubham Agarwal, Alexander Krentsel, et al.
    arXiv preprint 2026; ICML 2026 FAGEN Workshop Best Paper Oral
  5. arXiv'26
    Reality Is the Final Verifier: On Two Key Gaps in Agentic Software Engineering
    Alexander Krentsel*, Shubham Agarwal*, Mert Cemri*, Shu Liu*, Sidharth Sankhe, et al.
    arXiv preprint 2026
  6. arXiv'25
    Let the Barbarians In: How AI Can Accelerate Systems Performance Research
    Audrey Cheng*, Shu Liu*, Melissa Pan, Zhifei Li, Shubham Agarwal, Mert Cemri, et al.
    arXiv preprint 2025
  7. arXiv'26
    Dialogue to Discovery: Attribute-Aware Preference Elicitation for Conversational Product Search Assistants
    Sarthak Harne, Natwar Modani, Debabrata Mahapatra, Shubham Agarwal
    arXiv preprint 2026

Patents 8

  1. USPTO
    Using Reinforcement Learning to Recommend Data Visualizations
    Shubham Agarwal, Subrata Mitra, et al.
    US Patent 12,443,621 (granted)
  2. USPTO
    Data Exploration using Natural Language with Data Sampling
    Subrata Mitra, Shubham Agarwal, Gromit Yeuk-Yin Chan, Shaddy Garg, Tong Yu
    US Patent 12,664,160 (granted)
  3. USPTO
    Multi-Modal Retrieval using an Intermediate Noise State
    Shubham Agarwal, Subrata Mitra, et al.
    US Patent 12,694,578 (granted)
  4. USPTO
    Relative Contextualization using Pre-Softmax Logits in Transformers
    Debabrata Mahapatra, Shubham Agarwal, et al.
    US Patent Application 19/285,322
  5. USPTO
    Managing Chunk Caches for Efficient Retrieval-Augmented Generation in LLMs
    Shubham Agarwal, Archit Gupta, Nirmal Joshua Kapu, Debabrata Mahapatra, Subrata Mitra, et al.
    US Patent Application 19/074,061
  6. USPTO
    ReCon: Acceleration for Text-to-Image Synthesis with Concept Trajectories
    Chen-Yi Lu, Shubham Agarwal, et al.
    US Patent Application 63/698,535
  7. USPTO
    Micromanaging Prompts for High-Throughput Text-to-Image Inference
    Shubham Agarwal, Subrata Mitra, Saud Iqbal
    US Patent Application 18/808,654
  8. USPTO
    A System and Method for Outage Forecasting
    Shaddy Garg, Shubham Agarwal, Sumit Bisht, Nikhil Sheoran, Chahat Jain, et al.
    US Patent Application 17/656,263

Talks 15

† joint talk with co-authors.

Dec 2026 Inductive Deductive SynthesisOral presentation, NeurIPS 2026, Sydney (upcoming)
Nov 2026 Perfmaxxing SGLang for Next-Gen Model Serving with SkyDiscoverSGLang Summit
Nov 2026 Scaling Swarms of AgentsMicrosoft Azure Research (Systems)
Nov 2026 Continuous Optimization of Inference Engines with Self-Evolving AgentsAMD × UC Berkeley
Nov 2026 Towards Continuous Optimization of LLM Inference Systems with SkyDiscoverInferact†
Nov 2026 SkyDiscover-SynthesizeBroadcom × Sky Computing Lab Research Review
Oct 2026 SkyDiscover-SynthesizeEPIC Data Lab Retreat, Berkeley
Oct 2026 SkyDiscover-OptimizeAMD, San Jose†
Oct 2026 Supporting Our AI Overlords: Redesigning Data Systems to be Agent-FirstFrontier Research Club†
Sep 2026 Agents for OptimizationGoogle
Jun 2026 SkyDiscover: A Flexible Framework for AI-Driven Scientific and Algorithmic DiscoveryAIDDA 2026†
Apr 2026 Arming Data Agents with Tribal KnowledgeUC Berkeley Database Seminar
Nov 2024 Approximate Caching for Efficiently Serving Text-to-Image Diffusion ModelsUW Systems Group
Apr 2024 Approximate Caching for Efficiently Serving Text-to-Image Diffusion ModelsNSDI 2024, Santa Clara
Sep 2023 ESRO: Experience Assisted Service Reliability against OutagesASE 2023, Luxembourg

Honors and Service

  • Best Paper Award, ICML 2026 FAGEN Workshop, for Fantastic Adaptive Taxonomies and How to Use Them
  • UC Berkeley EECS Ph.D. Fellowship (2025)
  • Laude Institute Slingshots grant (SkyDiscover team, 2026)
  • Co-organizer and Area Chair, AI Agents for Discovery in the Wild workshop, CAIS 2026
  • External reviewer for NeurIPS, ICLR, ATC, AISTATS and AAAI
  • Officer, Computer Science Graduate Entrepreneurs (CSGE), UC Berkeley, helping turn research into startups
  • Industry mentor to a UMass master's student team researching LLM-based RAG systems