About Me

I'm an AI Engineer at Together AI, working on finetuning and distributed training of LLMs, building large scale GPU systems, and pushing the boundaries of Generative AI. I've worked in the field of generative modeling for over 8 years, largely focused on speech processing, computational music, and natural language processing. I graduated from University of California, Berkeley with a Bachelors of Science in Electrical Engineering & Computer Science in 2020, focused on Machine Learning. In the past I worked at Google, Berkeley AI Research, and Brilliant Inc. In my free time, I produce music, sketch, and game.

Contact Details

Due to privacy concerns, please contact me on LinkedIn for additional contact information.

Education

Stanford University

Non-Degree Sep '21 - Present

University of California, Berkeley

B.S. Electrical Engineering and Computer Science Aug '16 - May '20

Work

Together AI | Engineering

AI Engineer July '23 - Current

  • Built distributed systems with 6000+ GPUs for inference and training with PyTorch, DeepSpeed, FSDP, Golang, Slurm and Kubernetes.
  • Developed API training platform for 30+ LLMs with up to 256B tokens.
  • Research in parallel speculative decoding - upto 40% improvement.

Tikok | Speech Audio Music Intelligence

Research Engineer April '22 – April '23

  • Led research on voice beautification, attribute conversion, and voice feature dissentanglement.
  • Built experimentation pipeline for training 4 different large voice models with data processing on 3 different datasets and 40k speakers.
  • Developed voice design pipeline with zero-shot voice conversion, including age, gender and speaker interpolation.

NVIDIA | AI Applications

Deep Learning Engineer July '20 – Feb '22

  • Deployed model conversion tool for Riva Speech Services to optimize models for server deployment with Triton using ONNX and TensorRT. Supports over 15 different pipelines, accelerating for >12x.
  • Designed and built TAO-LM, tool for training/tuning N-Gram models.
  • Standardized testing framework to increase coverage from 40% to 72%.
  • Contributed to online demo for Riva, used by >500 users a day.

Berkeley AI Research | AUTOLab

Researcher & Grader Jan '19 - Jan '20

  • Research in Reinforcement, Imitation and Online Learning.
  • Reduced failure of safety using uncertainty estimation by 14%.
  • Built 8 experiments for imitation learning with an improving supervisor.
  • Graded/tutored for Deep Learning and Optimization for >600 students.

Google | Chrome Media Audio

Software Engineer Intern May '18- Aug '18

  • Created TF Estimators experimentation framework to predict the speech coding quality of WaveNet/Lyra while reducing bitrate by 50%.
  • Collected 7000 user-rated WaveNet samples. Ran experiments with RNNs, Dilated Convolutions and Variational Autoencoders.

Launchpad

President & Project Lead May '17- Dec'19

  • As president, led and organized educational ML workshops and meetings for 40+ members. Maintained relationships with 3 sponsors.
  • As PL, led 16 developers on 2 research-oriented projects.

Brilliant

Course Author May '17- Aug'18

  • Wrote tutorials, quizzes, and course content for Machine Learning and Advanced Algorithms courses.

Research

“On-Policy Robot Imitation Learning from a Converging Supervisor”

Ashwin Balakrishna*, Brijen Thananjeyan*, Jonathan Lee, Arsh Zahed, Felix Li, Joseph E. Gonzalez, Ken Goldberg

Conference on Robot Learning (CoRL), 2019. Oral Presentation. PDF

“On-Policy Imitation Learning from an Improving Supervisor”

Ashwin Balakrishna*, Brijen Thananjeyan*, Jonathan Lee, Arsh Zahed, Felix Li, Joseph E. Gonzalez, Ken Goldberg

Real World Sequential Decision Making Workshop at the International Conference on Machine Learning (ICML), 2019. PDF

Skills

My work cetralizes around a combination of Deep Learning, Speech Processing, and Reinforcement Learning. In the past, I've worked on proejcts involving Computational Music, NLP, Text-To-Speech, Model Based Reinforcement Learning, and Dynamics Estimation. I'm fluent in TensorFlow, PyTorch, and various other ML frameworks.

  • Deep Learning
  • Speech Processing
  • Natural Language Processing
  • Computational Music
  • Reinforcement Learning
  • TensorFlow
  • Pytorch

I also produce music.

I like to put my digital signal processing skills to a more creative use with electronic music production. I mostly produce for myself, so I haven't published much of my works yet, but here's an inside look at some of the projects I'm proud of or working on.

Get In Touch.

If you find something I've worked on interesting, feel free to get in touch. Please contact me through LinkedIn if you do not already have my contact info. I look forward to working with you!