PeopleofSuperintelligence

Researcher

Ankit Anand

Ankit Anand is a research scientist at Google DeepMind in Montreal whose work connects reasoning, reinforcement learning and mathematics. During his PhD at IIT Delhi, he developed symmetry-aware methods for UCT and probabilistic models. In 2020 and 2021, he coauthored work on machine-learned theorem proving. In 2024, he coauthored research on AlphaZero search for extremal graphs and code-generated reinforcement-learning rewards, and contributed to LearnLM for education. As of October 2026, he is also adjunct faculty at McGill University and an associate industry member at Mila.

Badges

Timeline

  1. 2015

    ASAP-UCT for tree search

    At IJCAI 2015, he coauthors ASAP-UCT with Aditya Grover, Mausam and Parag Singla. It brings ASAP-style state and state-action abstractions into UCT and improves policy quality by up to 26% on benchmark domains.

  2. 2019

    A PhD at IIT Delhi

    He completes a PhD at IIT Delhi in 2019 under Mausam and Parag Singla, with the dissertation Symmetry Aware Inference and Decision Making in Probabilistic Models. His doctoral work makes algorithms for probabilistic graphical models and Monte Carlo Tree Search symmetry-aware.

  3. 2020

    Synthetic theorems for proof learning

    In 2020, he coauthors Learning to Prove from Synthetic Theorems with Eser Aygün, Zafarali Ahmed, Doina Precup, Shibl Mourad and others. The paper trains an automated prover on synthetic theorems generated from axioms; the prover transfers to human-generated theorems and solves a substantial fraction of TPTP problems.

  4. 2021–2022

    Hindsight replay for theorem proving

    In 2021, he coauthors Proving Theorems using Incremental Learning and Hindsight Experience Replay, which uses intermediate data from unsuccessful proof attempts to train a prover. The paper appears at ICML 2022.

  5. 2024

    Extremal graphs with AlphaZero

    He coauthors Finding Increasingly Large Extremal Graphs with AlphaZero and Tabu Search for IJCAI 2024, listed second among its three joint first authors. The search finds edge-rich graphs without 3- or 4-cycles and improves lower bounds for several sizes on a problem inspired by a 1975 conjecture of Erdős.

  6. 2024

    Code as Reward at ICML

    He coauthors the ICML 2024 spotlight paper Code as Reward, which uses vision-language models to generate dense reinforcement-learning rewards and reduce the cost of repeated model queries. In May 2024, he gives a research talk at the University of Bonn on applying reinforcement learning to theorem proving and bringing formal code back to reinforcement learning.

  7. 2024

    LearnLM for education

    He contributes to LearnLM, generative AI work for teaching announced at Google I/O in 2024. His paper Towards Responsible Development of Generative AI for Education introduces LearnLM-Tutor, a Gemini model fine-tuned for pedagogy that educators and learners consistently prefer to a prompt-tuned Gemini.

  8. As of October 2026

    Research scientist at DeepMind

    As of October 2026, he is a research scientist at Google DeepMind in Montreal, an adjunct faculty member at McGill University and an associate industry member at Mila. His research links reasoning and reinforcement learning with mathematics and theoretical computer science.

Career

  1. NowGoogle DeepMindResearch Scientist
  2. NowMilaAssociate Industry Member
  3. NowMcGill UniversityAdjunct Faculty Member

Education

  1. 2019IIT DelhiPhD

From the biography.

  1. 16% of AI career moves in the 2020s went from a university to a companyIncludes Ankit Anand. We estimate that 16% of AI career moves in the 2020s went from a university to a company; 29% of moves starting at a university went to a company, and 76% ended the university role.
  2. Google had the most people shaping AI every year from 2014 to 2025Includes Google. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  3. Google has the highest combined h-index among AI organizationsIncludes Google. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.
  4. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Google. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Ankit Anand yet. Send us a free-licensed photo.

  1. ASAP-UCT: Abstraction of State-Action Pairs in UCT | ML Anthology mlanthology.org. Checked 22 October 2026.
  2. Mausam : Students cse.iitd.ac.in. Checked 22 October 2026.
  3. Ankit Anand | Mila mila.quebec. Checked 22 October 2026.
  4. [2006.11259] Learning to Prove from Synthetic Theorems arxiv.org. Checked 22 October 2026.
  5. [2112.10664] Proving Theorems using Incremental Learning and Hindsight Experience Replay arxiv.org. Checked 22 October 2026.
  6. Proving Theorems using Incremental Learning and Hindsight Experience Replay proceedings.mlr.press. Checked 22 October 2026.
  7. [2311.03583] Finding Increasingly Large Extremal Graphs with AlphaZero and Tabu Search arxiv.org. Checked 22 October 2026.
  8. ICML Poster Code as Reward: Empowering Reinforcement Learning with VLMs ICML 2024 icml.cc. Checked 22 October 2026.
  9. [2402.04764] Code as Reward: Empowering Reinforcement Learning with VLMs arxiv.org. Checked 22 October 2026.
  10. Colloquium: Research Talk by Dr. Ankit Anand — Institute of Computer Science informatik.uni-bonn.de. Checked 22 October 2026.
  11. Ankit Anand sites.google.com. Checked 22 October 2026.
  12. [2407.12687] Towards Responsible Development of Generative AI for Education: An Evaluation-Driven Approach arxiv.org. Checked 22 October 2026.