PeopleofSuperintelligence

Builder

Yash Lara

Yash Lara is a Senior PM Lead in Microsoft Research's AI Frontiers lab whose work spans synthetic-data generation, reasoning models and computer-use agents. They coauthored AgentInstruct and Orca-AgentInstruct in 2024, then Phi-4-reasoning and Fara-7B in 2025. In 2026, they coauthored Fara-1.5 and Echoverse, on computer-use agents and the environments used to train them. As of October 2026, they work with the Modelling and Evaluation team on AI Frontiers strategy, execution and model releases.

Links
LinkedIn

Timeline

  1. July 2024

    AgentInstruct and Orca-3

    They are among the 14 authors of AgentInstruct: Toward Generative Teaching with Agentic Flows, submitted in July 2024. The framework creates a 25-million-pair post-training dataset and uses it to fine-tune Mistral-7B into Orca-3.

  2. November 2024

    Orca-AgentInstruct release

    They coauthor the Microsoft Research post Orca-AgentInstruct: Agentic flows can be effective synthetic-data generators with Arindam Mitra and Ahmed Awadallah in November 2024. The work extends Orca and Orca 2's synthetic-data training with agentic flows, and releases a one-million-pair subset of the dataset.

  3. March 2025

    Inference-time scaling report

    They coauthor the Microsoft technical report MSR-TR-2025-16, Inference-Time Scaling for Complex Tasks: Where We Stand and What Lies Ahead, published in March 2025.

  4. April 2025

    The Phi-4-reasoning report

    In April 2025, they coauthor Phi-4-reasoning Technical Report, which introduces a 14-billion-parameter reasoning model. The model is trained by supervised fine-tuning Phi-4 on teachable prompts and reasoning demonstrations generated with o3-mini.

  5. November 2025

    Fara-7B for computer use

    They coauthor Fara-7B: An Efficient Agentic Model for Computer Use, which Microsoft announces in November 2025 as its first agentic small language model designed for computer use. The 7-billion-parameter model runs directly on devices and is released open-weight on Microsoft Foundry and HuggingFace.

  6. June 2026

    The Fara-1.5 model family

    They coauthor Fara-1.5 in June 2026, a family of native computer-use agents in 4B, 9B and 27B sizes trained with data from the FaraGen1.5 pipeline.

  7. July 2026

    Echoverse training environments

    In July 2026, they coauthor Echoverse: Deep, Evolving Environments for Training Computer-Use Agents at Scale. Its co-evolution loop uses graded agent rollouts to repair the synthetic environments, tasks and verifiers.

  8. As of October 2026

    Senior PM Lead at AI Frontiers

    They are a Senior PM Lead in the AI Frontiers lab at Microsoft Research and work with its Modelling and Evaluation team on strategy, execution and model releases. Their work translates research insights into product innovations.

  1. Among employers with 30+ full-time cases, OpenAI leads primary technical role H-1B pay at $310,000, and Amazon, Google and Microsoft file 57% of all casesIncludes Microsoft. Across 440,669 certified US H-1B filings from 44 employers (FY2020–FY2026), OpenAI leads primary technical role base pay at $310,000 and xAI leads overall base pay at $290,000.
  2. Google had the most people shaping AI every year from 2014 to 2025Includes Microsoft. 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. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Microsoft. 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.
  4. Google has the highest combined h-index among AI organizationsIncludes Microsoft. 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.

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 Yash Lara yet. Send us a free-licensed photo.

  1. AgentInstruct: Toward Generative Teaching with Agentic Flows arxiv.org. Checked 21 October 2026.
  2. Paper page - AgentInstruct: Toward Generative Teaching with Agentic Flows huggingface.co. Checked 21 October 2026.
  3. Orca-AgentInstruct: Agentic flows can be effective synthetic-data generators microsoft.com. Checked 21 October 2026.
  4. Yash Lara - Publications - Microsoft Research microsoft.com. Checked 21 October 2026.
  5. Phi-4-reasoning Technical Report arxiv.org. Checked 21 October 2026.
  6. Fara-7B: An Efficient Agentic Model for Computer Use - Microsoft Research Blog microsoft.com. Checked 21 October 2026.
  7. Fara-7B: An Efficient Agentic Model for Computer Use arxiv.org. Checked 21 October 2026.
  8. Fara-1.5: Scalable Learning Environments for Computer Use Agents arxiv.org. Checked 21 October 2026.
  9. Echoverse: Deep, Evolving Environments for Training Computer-Use Agents at Scale - Microsoft Research microsoft.com. Checked 21 October 2026.
  10. Yash Lara at Microsoft Research microsoft.com. Checked 21 October 2026.