Software Engineer: ML Optimization

Generalist
Boston, US
On-site

Who this role is best for

Geared toward candidates comfortable with optimizing both training and inference stacks in ML, and working in a collaborative, cutting-edge robotics and AI environment in Boston.

Best fit for

  • Candidates with deep expertise in optimizing transformer and diffusion models for performance
    — “proficient and stay current with the latest ML techniques for training and inference optimizations in transformer and diffusion based architectures”
  • Candidates with experience in GPU optimization and low-level system bottlenecks
    — “Making GPUs go brrrrr”

Things to consider

  • The role demands a high level of technical versatility across multiple domains.
    — “chase ML optimizations anywhere: From the CUDA kernels, to ML architecture, to frontend or backend network bottlenecks”

How to stand out

  • Highlight experience with end-to-end optimization of ML pipelines and hardware integration.
    — “Optimizing both the inference and training stacks”
  • Emphasize projects involving performance tuning of deep learning models in production.
    — “Implementing ML, hardware, and software changes that lead to step-function gains”
  • Demonstrate familiarity with low-level system components like CUDA and NVLink.
    — “chase ML optimizations anywhere: From the CUDA kernels, to ML architecture”
  • Showcase a history of driving breakthroughs in AI or robotics research or development.
    — “team of researchers, roboticists, and company builders come from OpenAI, Boston Dynamics, Google DeepMind”
Pace · Fast PacedCollaboration · HighAutonomy · HighDecision Impact · Team

Derived from job-description analysis by Serendipath's career intelligence engine.

What success looks like

  • implementing ML optimizations
  • optimizing training and inference stacks
Typical background
proficiency in ML techniques for training and inference optimizations

Skills & requirements

Required

ML OptimizationTraining And Inference Stack OptimizationCUDA Kernel OptimizationHardware And Software Changes

Preferred

Large Embodied Multimodal ModelsVision-language-action Models

Stack & domain

ML OptimizationTraining And Inference OptimizationsTransformer And Diffusion Based ArchitecturesCUDA KernelsML ArchitectureFrontend Or Backend Network BottlenecksCPU BottlenecksNvlink And CommsTorchNumPyPython InefficienciesLarge-scale AIRoboticsDataModelsHardwareDexterityEmbodied Foundation ModelsMultimodal ModelsVision-language-action ModelsAutonomous DrivingNext-generation RobotsLarge Embodied Multimodal ModelsProblem-solvingInnovationAdaptabilityTeamworkLeadershipCollaborationCommunicationProject ManagementContinuous LearningTechnical ProficiencySoftware EngineeringHardware EngineeringData EngineeringModel EngineeringRobotics EngineeringAI EngineeringAI

About the role

Original posting from Generalist via Ashby

ABOUT GENERALIST

At Generalist, we are on a mission to build general intelligence for the physical world and make it useful to everyone. We believe the industries and homes of the future will depend on humans and machines working together in new ways. Robots can help us build more and get more done.

We build embodied foundation models, starting with a focus on dexterity. This requires advancing the frontiers of data, models, and hardware, to enable robots to intelligently interact with the physical world.

The company embraces both large-scale AI and robotics as core to its DNA. Our team of researchers, roboticists, and company builders come from OpenAI, Boston Dynamics, Google DeepMind, and other frontier labs—with a track record of shipping AI breakthroughs. Before Generalist, we pioneered large embodied multimodal models and vision-language-action models (PaLM-E, https://research.google/blog/palm-e-an-embodied-multimodal-language-model/ RT-2 https://deepmind.google/blog/rt-2-new-model-translates-vision-and-language-into-action/, Gemini Robotics https://deepmind.google/models/gemini-robotics/), launched and scaled ChatGPT https://chatgpt.com/ and GPT-4 https://openai.com/index/gpt-4-research/ to hundreds of millions of users, engineered the foundations of autonomous driving, built next-generation robots (Atlas https://bostondynamics.com/atlas/, Spot https://bostondynamics.com/products/spot/, Stretch https://bostondynamics.com/products/stretch/) and pushed the limits of what they can do (from parkour https://www.youtube.com/watch?v=tF4DML7FIWk to manipulation https://bostondynamics.com/blog/large-behavior-models-atlas-find-new-footing/, and testing robustness https://www.youtube.com/watch?v=aFuA50H9uek).

We are an equal opportunity employer, and we do not discriminate on the basis of race, religion, color, national origin, sex, sexual orientation, age, veteran status, disability, genetic information, or other applicable legally protected characteristic.

ABOUT THE ROLE

We internally call this team MBMB (More Big More Better). You will own optimizations on both the training and on-robot inference stacks. We are still in a regime of step-function, not incremental, gains.

You’ll be responsible for:

  • Making GPUs go brrrrr
  • Implementing ML, hardware, and software changes that lead to step-function gains
  • Optimizing both the inference and training stacks

You might thrive in this role if you:

  • Are proficient and stay current with the latest ML techniques for training and inference optimizations in transformer and diffusion based architectures
  • Will chase ML optimizations anywhere: From the CUDA kernels, to ML architecture, to frontend or backend network bottlenecks, CPU bottlenecks, NVLink and comms, to torch, numpy, and Python inefficiencies.

Source: Generalist careers (Ashby)

Similar roles

More roles at Generalist