AI Researcher

Sara Hooker Portfolio

Co-founder of Adaption Labs, which is "betting against scaling" to build efficient AI that continually learns; previously VP of Research at Cohere leading Cohere Labs, and before that a research scientist at Google DeepMind; her site collects papers and essays on efficient AI and scaling limits, including "The Hardware Lottery" and "On the Slow Death of Scaling."

Efficient AIModel CompressionAI Research StrategyCompute Scaling

Location

Location: classified. Send a pigeon.

Companies

Adaption LabsCohereGoogle DeepMind

Credentials

  • PhD in Computer Science, Mila AI Institute (supervised by Hugo Larochelle and Aaron Courville)

Skills

Efficient AIModel CompressionAI Research StrategyCompute Scaling

Notable

  • Named to the TIME100 AI list (2024)
  • "The Hardware Lottery" published in Communications of the ACM (December 2021)
  • Co-founded Adaption Labs, which raised a $50M seed round in 2026 (Fortune)
  • Member of Kaggle's ML Advisory Research Board and the World Economic Forum council on the Future of AI (per her site)

What makes it work

A breakdown of the choices that make this portfolio stand out.

1

Named, quotable arguments instead of neutral paper titles

"The Hardware Lottery" and "The Leaderboard Illusion" each state a specific, memorable thesis, which is why they get cited and discussed well beyond their original academic venue.

2

A research agenda built around a specific counter-narrative

The homepage states the bet outright ("We are betting against scaling"), and the writing it lists (The Hardware Lottery, the limits of compute thresholds, the slow death of scaling) backs it up, a coherent and differentiated research position rather than a scattershot agenda.

3

External recognition treated as evidence, not decoration

Citing recognitions like the Time 100 AI list alongside the actual papers gives visitors an independent, checkable signal of the work's field-level impact.

What AI Researchers can take from this

Specific, actionable tips to apply to your own portfolio — no generic advice.

  • Give your papers or arguments a clear, quotable name if the underlying claim is genuinely a distinct thesis — it's what makes work travel beyond its original venue.

  • Build your research identity around a specific, stated bet (efficiency over scale, for instance) rather than covering a field generically.

  • List independent recognitions alongside your papers as one more piece of checkable evidence, not a replacement for the work itself.

Projects

The Hardware Lottery, The Leaderboard Illusion, the slow death of scaling: Sara Hooker names each paper like a thesis the reader is invited to argue with, and then founded a company on the argument.

ResearchadvancedSingle essay

The Hardware Lottery

HardwareML SystemsResearch History

A single-author essay arguing that research ideas win or lose partly on whether they fit the available hardware and software, not only on merit.

A named, portable thesis that frames how a field thinks about its own progress.

  • Single author

  • Also on arXiv (2009.06489)

  • Grounds her later efficiency-over-scale position

ResearchadvancedMulti-year research program

The Leaderboard Illusion

LLM EvaluationBenchmarks

A 2025 paper on how leaderboard practices can distort rankings of large language models.

Evaluation research with direct consequences for how labs report results.

  • Linked from 'Other writings'

  • Press coverage linked on-site

  • Fits her pattern of quotable titles

ResearchadvancedSingle essay

On the Slow Death of Scaling

Scaling LawsEfficient AI

An essay arguing against brute-force scaling as the main driver of AI progress, the same bet her company is built on.

Shows research positions carried through into what she builds next.

  • Most recent essay on the homepage

  • Mirrors the Adaption pitch

  • Continues The Hardware Lottery's line of argument

Ready to build your portfolio?

Follow the AI Researcher roadmap — skills, projects, and timeline to get hired.

AI Researcher Roadmap