PeopleofSuperintelligence

Researcher

Megha Srivastava

Megha Srivastava is a computer scientist whose research connects machine learning with fairness, adaptive education and human-AI collaboration. She coauthored a 2017 study of visual invariance in convolutional networks and a 2018 ICML best paper runner-up on fairness without demographic data. Her later work covered adaptive question generation, AI code security, verifiable training, motor-skill teaching and emergent misalignment after narrow language-model fine-tuning. She is a final-year Stanford computer science PhD candidate and an AI Science Writer at Anthropic.

Now at
Anthropic
Badges

Timeline

  1. 2017

    Visual invariance in CNNs

    She coauthors Learning Strategy versus Inherent Architecture on the Ability of CNNs to Develop Transformation Invariance with Kalanit Grill-Spector for the Vision Sciences Society in 2017.

  2. 2018

    Fairness without demographics

    She coauthors Fairness Without Demographics in Repeated Loss Minimization. The paper applies distributionally robust optimization to control risk for minority groups while remaining oblivious to group identities, and ICML names it a best paper runner-up in 2018.

  3. 2019

    A Stanford degree and residency

    After earning a Stanford B.S. in Computer Science and a Creative Writing minor, she serves as an AI Resident at Microsoft Research Redmond in 2019. She works with the Adaptive Systems & Interaction group on backward compatibility in machine-learning systems and plans to return to Stanford for a Computer Science PhD.

  4. 2021

    Questions for adaptive education

    In 2021, she and Noah Goodman publish Question Generation for Adaptive Education, using pretrained language models to estimate student knowledge and generate questions for a target difficulty.

  5. 2022–2024

    AI code security and training

    In 2022, she coauthors Do Users Write More Insecure Code with AI Assistants?, a study that finds participants with access to an AI code assistant write less secure code than those without. In 2024, she coauthors Optimistic Verifiable Training by Controlling Hardware Nondeterminism, which enables exact replication of model training across GPUs.

  6. 2025

    Teaching through shared autonomy

    In 2025, she coauthors Shared Autonomy for Proximal Teaching, which uses shared autonomy to give personalized instruction for motor skills. A 50-person simulated racing study finds gains in driving time, behavior and smoothness.

  7. 2026

    A Nature paper on misalignment

    In January 2026, she coauthors a Nature paper on emergent misalignment. It finds that fine-tuning a language model on a narrow insecure-code task can trigger concerning behavior beyond coding.

  8. As of October 2026

    Stanford PhD and Anthropic role

    As of October 2026, she is a final-year PhD candidate in Computer Science at Stanford, co-advised by Dan Boneh and Dorsa Sadigh, and an AI Science Writer at Anthropic. Her research includes interactive AI for education and scientific discovery, plus machine-learning security and verifiable training. In October 2026, she presents the Proximal State Nudging preprint with Toyota at Stanford's Center for AI Safety Annual Meeting; the work studies reducing skill atrophy from AI assistance.

Career

  1. NowAnthropicAI Science Writer
  2. 2019Microsoft ResearchAI Resident

Education

  1. Stanford UniversityBS in Computer Science

From the biography.

  1. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Megha Srivastava. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.
  2. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Anthropic. Our estimates of where OpenAI recruits came from and where 133 people with former OpenAI roles work now, including the companies and labs they founded.
  3. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Anthropic. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  4. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Anthropic. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.

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 Megha Srivastava yet. Send us a free-licensed photo.

  1. Research meghabyte.github.io. Checked 21 October 2026.
  2. Fairness Without Demographics in Repeated Loss Minimization proceedings.mlr.press. Checked 21 October 2026.
  3. Tatsunori Hashimoto | Home thashim.github.io. Checked 21 October 2026.
  4. AI Residency: Year Two | Meet Megha - Writer, Photographer, Computer Scientist, and AI Resident - Microsoft Research microsoft.com. Checked 21 October 2026.
  5. Question Generation for Adaptive Education - ACL Anthology aclanthology.org. Checked 21 October 2026.
  6. Do Users Write More Insecure Code with AI Assistants? arxiv.org. Checked 21 October 2026.
  7. Optimistic Verifiable Training by Controlling Hardware Nondeterminism papers.nips.cc. Checked 21 October 2026.
  8. Shared Autonomy for Proximal Teaching arxiv.org. Checked 21 October 2026.
  9. Training large language models on narrow tasks can lead to broad misalignment nature.com. Checked 21 October 2026.
  10. Megha Srivastava meghabyte.github.io. Checked 21 October 2026.