PeopleofSuperintelligence

Researcher

Olli Saarikivi

Olli Saarikivi is a machine learning researcher whose work spans encrypted computation, GPU communication and language models. He earned computer science degrees in Finland, coauthored CHET and EVA for homomorphic computation, and built MSCCL, a communication library for AI workloads. At Microsoft, he contributed to the Phi family through April 2025, then joined Apple’s Foundation Models team; his MSCCL++ paper received an ASPLOS 2026 Best Paper honorable mention. Since June 2026, he is an AI research scientist at Meta, working on pretraining data and large-scale synthetic generations for the Muse frontier models.

Now at
Meta
Works
52
Citations
974
h-index
15
Badges

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. 2007–2018

    Degrees and symbolic transducers

    He studies computer science at Helsinki University's Department of Computer Science from 2007 to 2011 and earns a Bachelor of Science; at Aalto University's Department of Computer Science, he earns a Master of Science (Technology) in 2013 and a Doctor of Science (Technology) in 2018. In 2017, he and Margus Veanes publish “Minimization of Symbolic Transducers” at CAV; the paper presents an algorithm for computing quotients of symbolic transducers.

  2. 2018–2020

    Homomorphic encryption compilers

    He joins Microsoft as a postdoctoral researcher in April 2018. His CHET paper appears on arXiv in October 2018 and at PLDI in 2019; CHET is a compiler and runtime for evaluating neural-network models on encrypted data, and in August 2019 he is a postdoctoral researcher in the RiSE group. In 2020, he coauthors EVA, a language and compiler for homomorphic computation whose programs average 5.3 times faster than those generated by CHET.

  3. 2020–2024

    MSCCL communication systems

    From September 2020 to January 2024, he is a senior researcher at Microsoft and builds MSCCL, a programmable communication library that delivers significant speedups to OpenAI's inference workloads; he leads its language design and writes its compilers in Python and MLIR. In 2023, he coauthors MSCCLang (Microsoft Collective Communication Language) at ASPLOS, a domain-specific language and optimizing compiler for GPU collective algorithms; its AllReduce and AllToAll algorithms run up to 1.9 and 1.3 times faster, respectively, than hand-optimized implementations.

  4. 2023–2025

    Phi-1 and Phi-3

    In 2023, he coauthors “Textbooks Are All You Need,” which introduces phi-1, a 1.3-billion-parameter code model. In January 2024, he becomes a Member of Technical Staff at Microsoft, a role he holds through April 2025, and works on Phi small language models, generating synthetic data, filtering web data and running ablations. The 2024 Phi-3 report, which he coauthors, introduces phi-3-mini, a 3.8-billion-parameter model small enough to run on a phone.

  5. 2024–2025

    Phi-4 and reasoning

    In December 2024, he coauthors the Phi-4 technical report, which presents a 14-billion-parameter model trained with a focus on data quality and synthetic data. At Microsoft AI's pretraining team, he leads synthetic-data tool development and builds generations to improve phi-4 reasoning. In 2025, he coauthors Phi-4-reasoning, a 14-billion-parameter reasoning model that he distills from an OpenAI reasoning model.

  6. April 2025–May 2026

    Apple Foundation Models

    In April 2025, he moves from Microsoft to Apple as a machine learning researcher on Apple's Foundation Models team, improving STEM capabilities with pretraining-scale synthetic data and tuning data mixtures through May 2026. In 2025, he coauthors the Apple Intelligence Foundation Language Models report, which describes two multilingual, multimodal models that power Apple Intelligence features. In March 2026, he coauthors CHIMERA, a compact synthetic reasoning dataset of 9K samples designed to support generalizable reasoning across domains.

  7. 2026

    MSCCL++ at ASPLOS

    In March 2026, he coauthors MSCCL++, a communication system for AI inference. The ASPLOS paper receives a Best Paper honorable mention in 2026.

  8. Since June 2026

    AI research scientist at Meta

    In June 2026, he joins Meta as an AI research scientist, working on pretraining data and large-scale synthetic generations for the Muse frontier models.

Career

  1. Since 2026MetaAI research scientist
  2. 2025–2026AppleMachine learning researcher
  3. 2018–2025MicrosoftPostdoctoral researcher, Senior researcher

Education

  1. 2018Aalto UniversityDoctor of Science (Technology)
  2. 2013Aalto UniversityMaster of Science (Technology)
  3. 2011University of HelsinkiBachelor of Science

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

Most-cited works

  1. CHET: an optimizing compiler for fully-homomorphic neural-network inferencing2019 · Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation · 237 citationsWith Kristin E. Lauter and Madanlal Musuvathi
  2. Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone2024 · 164 citationsWith Adil Salim, Ahmed Awadallah, Ce Liu, Corby Rosset, Gustavo de Rosa and Harkirat Behl
  3. Textbooks Are All You Need2023 · 138 citationsWith Adam Tauman Kalai, Adil Salim, Gustavo de Rosa, Harkirat Behl, Mojan Javaheripi and Piero Kauffmann
  4. Breaking the computation and communication abstraction barrier in distributed machine learning workloads2022 · 69 citationsWith Madanlal Musuvathi
  5. MSCCLang: Microsoft Collective Communication Language2023 · ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) · 40 citationsWith Madanlal Musuvathi

From OpenAlex, as of 21 October 2026.

  1. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Olli Saarikivi. 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. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Meta. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  3. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Meta. 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.
  4. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Meta. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

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

  1. Olli Saarikivi / CV olli.saarikivi.fi. Checked 21 October 2026.
  2. Minimization of Symbolic Transducers - Microsoft Research microsoft.com. Checked 21 October 2026.
  3. CHET: Compiler and Runtime for Homomorphic Evaluation of Tensor Programs arxiv.org. Checked 21 October 2026.
  4. HE compilers for Private AI and other game changers with Dr. Olli Saarikivi - Microsoft Research microsoft.com. Checked 21 October 2026.
  5. EVA: An Encrypted Vector Arithmetic Language and Compiler for Efficient Homomorphic Computation arxiv.org. Checked 21 October 2026.
  6. ASPLOS 2023: Proceedings of the 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2 sigplan.github.io. Checked 21 October 2026.
  7. Textbooks Are All You Need arxiv.org. Checked 21 October 2026.
  8. Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone arxiv.org. Checked 21 October 2026.
  9. Phi-4 Technical Report - Microsoft Research microsoft.com. Checked 21 October 2026.
  10. Phi-4 Technical Report arxiv.org. Checked 21 October 2026.
  11. Phi-4-reasoning Technical Report arxiv.org. Checked 21 October 2026.
  12. Apple Intelligence Foundation Language Models: Tech Report 2025 arxiv.org. Checked 21 October 2026.
  13. CHIMERA: Compact Synthetic Data for Generalizable LLM Reasoning arxiv.org. Checked 21 October 2026.
  14. MSCCL++: Rethinking GPU Communication Abstractions for AI Inference - Microsoft Research microsoft.com. Checked 21 October 2026.