Michael Sachs

Summary

AI systems leader with 15+ years building and scaling machine learning and frontier AI capabilities across research, platform, and product domains. Proven track record translating cutting-edge machine learning — including large language models, deep learning systems, and large-scale experimentation platforms — into production systems used by millions of users and global enterprises.

Experienced leading interdisciplinary teams of researchers, engineers, and product leaders to operationalize advanced AI in complex environments. Deep expertise in LLM post-training, ML infrastructure, distributed systems, and enterprise AI deployment.

Selected AI Leadership Impact

  • Built and scaled applied machine learning organizations delivering production AI systems across global infrastructure and enterprise AI deployments.
  • Led development of LLM post-training and customization pipelines enabling domain-specific model capabilities for large enterprise customers.
  • Founded Netflix’s ML-for-Systems research team, applying machine learning to optimize distributed cloud infrastructure at hyperscale.
  • Designed experimentation platforms enabling safe iteration and large-scale AI deployment across complex distributed systems.

Technical Expertise

Machine Learning & AI

Large Language Models • LLM Post-Training (RLHF, SFT, Preference Optimization) • Deep Learning Systems • Transformers • PyTorch • Causal Inference • Large-Scale Experimentation • Applied AI Systems

ML Infrastructure & Data Systems

Distributed ML Systems • Spark/PySpark • Databricks • Airflow • MLflow • DVC • AWS • GCP • BigQuery • Redshift

Programming

Python • SQL • JavaScript • C/C++

Experience

Cohere

Senior Manager of Technical Staff, Applied Machine Learning Sep 2025 - Present

Lead Applied Machine Learning initiatives focused on post-training and customization of frontier large language models for global enterprise deployments.

  • Lead an applied ML organization responsible for extending the capabilities of foundation models through post-training techniques including supervised fine-tuning, preference optimization, and evaluation systems.
  • Develop domain-specific AI capabilities across enterprise use cases including document understanding, structured generation, and knowledge extraction.
  • Partner with major global organizations — including LG, Fujitsu, Hanwha Ocean, RWS, Oracle, and Dell — to deploy customized AI systems built on frontier models.
  • Own technical stewardship for enterprise AI deployments, including model performance, inference efficiency, and infrastructure cost optimization.
  • Built organizational foundations and career ladders for machine learning engineers and modeling specialists during a phase of rapid team expansion.
  • Developed hiring strategies and role definitions that significantly increased the seniority and technical depth of the candidate pipeline.

Netflix

Senior Data Science Manager, Infrastructure Jan 2022 - Sept 2024

Led machine learning research and engineering initiatives applying ML to optimize global infrastructure systems powering Netflix’s streaming platform.

  • Founded and led Netflix’s ML-for-Systems team, applying machine learning to large-scale distributed infrastructure optimization.
  • Developed models for traffic prediction, workload placement, and infrastructure capacity planning across global cloud systems.
  • Built LLM-based semantic systems linking customer support signals with internal telemetry, accelerating infrastructure issue detection and resolution.
  • Developed causal inference and experimentation platforms enabling safe infrastructure iteration across large-scale distributed services.
  • Partnered with engineering and finance teams to deploy machine learning systems optimizing utilization across $800M+ annual AWS infrastructure spend.

Data Science Manager, Content Distribution Infrastructure Sept 2020 - Jan 2022

Led cross-functional teams developing machine learning systems improving performance and reliability of Netflix’s global content delivery infrastructure.

  • Architected ML systems for VPN detection, traffic forecasting, and proactive content placement across the Open Connect network.
  • Built experimentation infrastructure enabling large-scale performance experimentation on network and delivery systems.
  • Spearheaded data infrastructure supporting Netflix Cloud Games, enabling telemetry collection and experimentation prior to launch.
  • Developed ultra-high-resolution TCP telemetry pipelines enabling new research into network performance optimization.

FLYR

Head of Product Mar 2019 - Apr 2020

Led a global cross-functional organization building AI-driven pricing and revenue optimization systems for major airline customers.

  • Directed a 22-person team spanning machine learning research, product management, and engineering across three continents.
  • Delivered ML-driven pricing and forecasting systems improving revenue optimization and customer retention.

Head of ML Platform Nov 2018 - Mar 2019

Built core machine learning infrastructure supporting large-scale model development and deployment.

  • Designed ML platform using GCP, DVC, and distributed training systems that reduced model iteration cycles by 80%.
  • Built scalable inference systems reducing production latency by 50% and production errors by 96%.

Radius Intelligence

Data Science Manager Mar 2017 - Oct 2018

Led applied research team developing machine learning systems improving B2B entity resolution and data quality.

  • Developed clustering, entity resolution, and data enrichment models for large-scale marketing datasets.
  • Established research impact metrics and career frameworks improving collaboration across engineering and product teams.

Discovery Digital Networks

Director of Data Science and Technology Sept 2014 - Mar 2017

  • Led migration of analytics and data infrastructure to AWS, significantly improving engineering velocity.
    • .
  • Built real-time analytics pipelines processing 500M+ daily events supporting global digital video platforms.

Data Scientist Mar 2014 - Sept 2014

  • Founded the organization’s data science function.
  • Built forecasting and viewership models used in content planning and programming strategy.

Early Career

  • Graduate Researcher and NASA Earth & Space Science Fellow — computational physics research (2010–2013)
  • Product and design leadership roles at Weill Cornell Medical College and Xperts (1997–2005)

Education

University of California, Davis

PhD in Physics, 2013 (GPA: 3.87)

Research focused on computational modeling and large-scale scientific data analysis

Columbia University

Completed undergraduate physics curriculum (GPA: 3.99)

Virginia Commonwealth University

BFA, Graphic Design, 1995 (GPA: 3.24)

Publications

Parametrizing Physics-Based Earthquake Simulations
K. W. Schultz, M. R. Yoder, J. M. Wilson, E. M. Heien, M. K. Sachs, J. B. Rundle, and D. L. Turcotte
Pure and Applied Geophysics  (2016)

Virtual Quake: Statistics, Co-Seismic Deformations and Gravity Changes for Driven Earthquake Fault Systems
K. W. Schultz, M. K. Sachs, E. M. Heien, M. R. Yoder, J. B. Rundle, D. L. Turcotte, and A. Donnellan
International Symposium on Geodesy for Earthquake and Natural Hazards (GENAH)  145  29-37  (2015)

Simulating Gravity Changes in Topologically Realistic Driven Earthquake Fault Systems: First Results
K. W. Schultz, M. K. Sachs, E. M. Heien, J. B. Rundle, D. L. Turcotte, and A. Donnellan
Pure and Applied Geophysics  Volume 17 827–838 (2014)

Self-Organizing Complex Earthquakes: Scaling in Data, Models, and Forecasting
M. K. Sachs, J. B. Rundle, J. R. Holliday, J. Gran, M. Yoder and W. Graves
"Self-Organized Criticality Systems"  Open Academic Press  (2013)

A Comparison among Observations and Earthquake Simulator Results for the allcal2 California Fault Model
T. E. Tullis, K. Richards-Dinger, M. Barall, J. H. Dieterich, E. H. Field, E. M. Heien, L. H. Kellogg, F. Pollitz, J. B. Rundle, M. K. Sachs, D. L. Turcotte, S. N. Ward and M. B. Yikilmaz
Seismological Research Letters  83  994-1006  (2012)

Generic Earthquake Simulator
T. E. Tullis, K. Richards-Dinger, M. Barall, J. H. Dieterich, E. H. Field, E. M. Heien, L. H. Kellogg, F. Pollitz, J. B. Rundle, M. K. Sachs, D. L. Turcotte, S. N. Ward and M. B. Yikilmaz
Seismological Research Letters  83  959-963  (2012)

Virtual California Earthquake Simulator
M. K. Sachs, E. M. Heien, D. L. Turcotte, M. B. Yikilmaz, J. B. Rundle and L. H. Kellogg
Seismological Research Letters  83  973-978  (2012)

Forecasting Earthquakes: The RELM Test
M. K. Sachs, D. L. Turcotte, J. R. Holliday and J. B. Rundle
Computing in Science and Engineering  14  43  (2012)

Understanding Long-Term Earthquake Behavior through Simulation
E. M. Heien and M. K. Sachs
Computing in Science and Engineering  14  10  (2012)

Black swans, power laws, and dragon-kings: Earthquakes, volcanic eruptions, landslides, wildfires, floods, and SOC models
M. K. Sachs, M. R. Yoder, D. L. Turcotte, J. B. Rundle and B. D. Malamud
European Physical Journal Special Topics  205  167-182  (2012)

Implications of the RELM test of earthquake forecasts in California
M. K. Sachs, Y. T. Lee, D. L. Turcotte, J. R. Holliday and J. B. Rundle
Research in Geophysics  2  e10  (2012)

Evaluating the RELM test results
M. K. Sachs, Y. T. Lee, D. L. Turcotte, J. R. Holliday and J. B. Rundle
International Journal of Geophysics  2012  (2012)

Earthquake precursors: activation or quiescence?
J. B. Rundle, J. R. Holliday, M. Yoder, M. K. Sachs, A. Donnellan, D. L. Turcotte, K. F. Tiampo, W. Klein and L. H. Kellogg
Geophysical Journal International  187  225-236  (2011)

Results of the Regional Earthquake Likelihood Models (RELM) test of earthquake forecasts in California
Y. T. Lee, D. L. Turcotte, J. R. Holliday, M. K. Sachs, J. B. Rundle, C. C. Chen and K. F. Tiampo
Proceedings of the National Academy of Sciences (USA)  108  16533-16538  (2011)

Testing Lattice Quantum Gravity in 2+1 Dimensions
M. K. Sachs
arXiv:1110.6880 [gr-qc]  (2011)

Awards

  • Three time Discovery D-Lighter award winner
  • 2014 and 2015 Discovery Digital Networks Hackathon winner
  • 2011 NASA Earth and Space Science Fellowship
  • 2011 Santa Fe Institute Complex Systems Summer School
  • Member of the Golden Key International Honor Society
  • Interactive Best in Show, Richmond Ad Show 2004: AdCenter Website
  • 10 eHealthcare Leadership Awards including 2 Platinum Awards for work done on the Weill Cornell Medical College Environmental Geriatrics Continuing Medical Education Application 2002-2005
  • Artwork exhibited in the 2002 Paperveins Museum of Art Biennial at the Here Arts Center in New York City
  • Xperts Employee of the Month May 1999, June 2001, August 2001
  • Addy Award: Xperts Self Promotional Website 2000
  • Xperts Outstanding Engineering Sales Support December 1999
  • Xperts Excellence in Engineering Award August 1999
  • 9 Xperts customer service awards 1997-2002

Conferences

AGU 2013

Earthquake Simulations and Historical Patterns of Events: Forecasting the Next Great Earthquake in California
M. K. Sachs, J. B. Rundle, E. M. Heien, K. Schultz, D. L. Turcotte, M. B. Yikilmaz, and L. H. Kellogg (2013)
Abstract NG41A-1662 (Poster) presented at 2013 Fall Meeting AGU San Francisco, Calif. 7-13 Dec.

Monitoring Earthquake Fault Slip from Space: Model Implications for a High Precision, High Resolution Dedicated Gravity Mission (Invited)
J. B. Rundle, M. K. Sachs, K. F. Tiampo, J. Fernandez, D. L. Turcotte, A. Donnellan, E. M. Heien and L. H. Kellogg (2013)
Abstract G13C-08 presented at 2013 Fall Meeting AGU San Francisco, Calif. 7-13 Dec.

AGU 2012

Virtual California: studying earthquakes through simulation
M. K. Sachs, E. M. Heien, D. L. Turcotte, M. B. Yikilmaz, J. B. Rundle and L. H. Kellogg (2012)
Abstract NG43C-02 presented at 2012 Fall Meeting AGU San Francisco, Calif. 3-7 Dec.

Dynamics, Patterns, and Migration in Earthquake Fault Systems (Invited)
J. B. Rundle, M. K. Sachs, J. R. Holliday, E. M. Heien, D. L. Turcotte, A. Donnellan and Z. Meadows (2012)
Abstract S13A-2518 (Poster) presented at 2012 Fall Meeting AGU San Francisco, Calif. 3-7 Dec.

EcoSummit 2012

Using Insights from Statistical Physics to Model Common Pool Resource Management
M. K. Sachs, N. Kunz, Z. A. Hamstead, A. Fajardo (2012)
Abstract GS07.28 presented at 2012 Meeting EcoSummit Columbus, Ohio 30 Sept. - 5 Oct.

AOGS 2012

Delivery of Earthquake Forecasts on Web-Based Platforms: Estimating Reliability and Forecast Skill
J. B. Rundle, J. R. Holliday, M. K. Sachs, W. Graves, P. B. Rundle, S. N. Ward and A. Donnellan (2012)
Abstract SE61-75-A001 presented at 2012 Meeting AOGS Singapore 13-17 Aug.

Numerical Simulations for Space-time Seismic Pattern Analysis and Earthquake Forecasting
M. K. Sachs, E. M. Heien, D. L. Turcotte, M. B. Yikilmaz, J. B. Rundle and L. H. Kellogg (2012)
Abstract SE61-75-A002 presented at 2012 Meeting AOGS Singapore 13-17 Aug.

AGU 2011

RELM Test Results: How Good Were the Forecasts?
M. K. Sachs, Y. T. Lee, D. L. Turcotte, J. R. Holliday and J. B. Rundle (2011)
Abstract NG44B-02 presented at 2011 Fall Meeting AGU San Francisco, Calif. 5-9 Dec.

Using Speculative Execution to Reduce Communication in a Parallel Large Scale Earthquake Simulation
E. M. Heien, M. B. Yikilmaz, M. K. Sachs, J. B. Rundle, D. L. Turcotte and L. H. Kellogg (2011)
Abstract NG51D-1672 (Poster) presented at 2011 Fall Meeting AGU San Francisco, Calif. 5-9 Dec.

E-DECIDER: Earthquake Disaster Decision Support and Response Tools - Development and Experiences
M. T. Glasscoe, R. G. Blom, G. W. Bawden, G. Fox, M. Pierce, J. B. Rundle, J. Wang, Y. Ma, M. R. Yoder, M. K. Sachs and J. W. Parker (2011)
Abstract IN11A-1269 (Poster) presented at 2011 Fall Meeting AGU San Francisco, Calif. 5-9 Dec.

SCEC Earthquake Simulator Comparison Results for California (Invited)
T. E. Tullis, K. Richards-Dinger, M. Barall, J. H. Dieterich, E. H. Field, E. M. Heien, L. H. Kellogg, F. Pollitz, J. B. Rundle, M. K. Sachs, D. L. Turcotte, S. N. Ward and O. Zielke (2011)
Abstract NG44B-01 presented at 2011 Fall Meeting AGU San Francisco, Calif. 5-9 Dec.

SCEC 2011

An Evaluation of the RELM Test Forecasts
M. K. Sachs, Y. T. Lee, D. L. Turcotte, J. R. Holliday and J. B. Rundle (2011)
Abstract B-120 (Poster) presented at 2011 Annual Meeting SCEC Palm Springs, Calif. 11-14 Sep.

Parallelization of the Virtual California Earthquake Simulator
E. M. Heien, M. B. Yikilmaz, M. K. Sachs, J. B. Rundle, L. H. Kellogg, and D. L. Turcotte (2011)
Abstract B-087 (Poster) presented at 2011 Annual Meeting SCEC Palm Springs, Calif. 11-14 Sep.

The Future of Virtual California Simulations
M. B. Yikilmaz, J. B. Rundle, D. L. Turcotte, E. M. Heien, M. K. Sachs, and L. H. Kellogg (2011)
Abstract B-110 (Poster) presented at 2011 Annual Meeting SCEC Palm Springs, Calif. 11-14 Sep.

Comparisons Among Earthquake Simulator Results for UCERF2 Fault Model of California and Observed Seismicity
T. E. Tullis, K. Richards-Dinger, M. Barall, J. H. Dieterich, E. H. Field, E. Heien, L. H. Kellogg, F. Pollitz, J. B. Rundle, M. K. Sachs, D. L. Turcotte, S. N. Ward, M. B. Yikilmaz, and O. Zielke (2011)
Abstract B-109 (Poster) presented at 2011 Annual Meeting SCEC Palm Springs, Calif. 11-14 Sep.

ACES 2011

Virtual California: Inner Workings, Recent Results and Future Development
M. K. Sachs, J. B. Rundle, D. L. Turcotte, A. Donnellan and J. W. Parker (2011)
Abstract 7400 presented at 2011 Meeting ACES Maui, Hawaii 1-5 May

Virtual California: A Guided Tour
M. K. Sachs, E. M. Heien, J. B. Rundle, D. L. Turcotte, M. B. Yikilmaz, L. H. Kellogg, K. F. Tiampo, A. Donnellan, W. Klein and J. W. Parker (2011)
Presented at 2011 Meeting ACES Maui, Hawaii 1-5 May

Press

Scientific American

Test Pits Earthquake Forecasts against Each Other

NBC News

Flagging quake hotspots an inexact science

UC Davis News

Assessing California earthquake forecasts

NASA

Managing the Deluge of 'Big Data' From Space

QuakeSim and NASA Mobile App Win NASA Software Award