Jeff Smith

Timeline

1. By 2015

   Bioinformatics and Intent Media

   After building bioinformatics software to reason about genomic data, he shifts toward broader machine learning and joins Ben Goertzel's project to predict the Hong Kong stock market with AI. In June 2015, as a data engineer at Intent Media, he writes an AWS Big Data Blog guest post on large-scale machine learning with Spark on Amazon EMR.

2. By 2018

   AI assistants at x.ai and IPsoft

   After about two years as an AI architect at x.ai, he builds meeting assistants Amy and Andrew. In July 2017, he and Wesley Harris launch John Done, an AI telephone assistant that makes calls for its users. By February 2018, he manages IPsoft's conversational AI team building Amelia.

3. 2018–2019

   Machine Learning Systems

   Manning publishes his book Machine Learning Systems: Designs that scale in May 2018; it teaches reactive design through Spark, MLlib and Akka. At the intersection of functional programming, distributed systems and machine learning, he coins the term reactive machine learning for an ideal machine-learning architecture and its techniques. In April 2019, he selects chapters for Manning's free eBook Exploring Deep Learning for Language.

4. 2018–2019

   PyTorch engineering manager

   In 2018, as an engineering manager at Facebook AI Research, he gives a talk on PyTorch 1.0 as a platform for moving AI research into production. During his time leading the PyTorch core development team, he works on the Vision, Text and Audio libraries, distributed training, and TorchServe with Amazon and the community. In May 2019, he writes about model serving on the PyTorch blog, and by October 2019 he is the senior engineering manager supporting the PyTorch team at Facebook AI.

5. 2020

   PyTorch Distributed

   In June 2020, he coauthors PyTorch Distributed: Experiences on Accelerating Data Parallel Training with Shen Li, Soumith Chintala and other PyTorch engineers; the paper describes the design, implementation and evaluation of PyTorch's distributed data-parallel module. His research leadership at FAIR supports projects including CICERO, SEER, ESMFold, CrypTen and Galactica, and he helps bring Papers with Code to Facebook AI.

6. July 2025

   Neural wristbands at Reality Labs

   At Meta Reality Labs, he works on EMG-based neuromotor interfaces that launch as neural wristbands. Published on 23 July 2025, a Nature paper lists him among its CTRL-labs at Reality Labs authors and describes an sEMG wristband whose decoding models generalize across people.

7. August 2025

   Launching 2nd Set AI

   Founded by Smith and Saurav Pandit in 2024, 2nd Set AI formally launches in August 2025 with Smith as CEO and co-founder. Its platform helps entertainment and sports media companies create new images and videos from characters and other imagery they own or license.

8. As of October 2026

   C Proof and Chelis

   In August 2026, he introduces C Note, a browser notebook for pricing and risk models. His software projects include Chelis, a statically typed functional language for AI-generated numerical code; he also makes occasional angel investments in early-stage technology companies, primarily in AI. As of October 2026, he is CEO and co-founder of C Proof, a startup building verified numerical code intelligence.