Sr. Machine Learning Engineer, Physical AI

Greylock Partners; San Francisco (CA); On-site; USD 180,000 - 280,000; Full time

Benefits offered by this job

Remote work option

Direct employment with portfolio公司

Job summary

Greylock Partners is hiring a machine learning engineer to join a seed-stage portfolio company in New York building an agentic development platform for autonomous physical systems. This is not a traditional ML role, you’ll build models that help engineers understand, predict, and improve complex physical systems.

You will work at the intersection of ML, simulation, sensing, and platform engineering, deploying end-to-end ML workflows and reusable capabilities for customers.

Qualifications

• Production ML systems built with Python and frameworks such as PyTorch or JAX.
• Models from complex numerical, temporal, or sensor data.
• Experience with robotics, autonomy, or simulation domains.
• Surrogate/learned dynamics models and system identification methods.

Responsibilities

• Build ML models from simulation, experimental, sensor, and real-world data.
• Develop surrogate and learned models for expensive physical processes.
• Build systems for system identification, parameter estimation, calibration, and optimization.
• Develop methods to measure and reduce the gap between simulated and observed real-world behavior.
• Build ML-driven workflows for anomaly detection, validation, and continuous improvement.
• Develop data and training pipelines that combine simulated and real-world observations.
• Design evaluation frameworks for model accuracy, uncertainty, robustness, and generalization.
• Explore active learning, uncertainty estimation, optimization, and adaptive experimentation.
• Productionize models as reliable components of a broader engineering platform.
• Work with customers to generalize learnings into reusable ML and platform capabilities.
• Help define the ML architecture and technical direction of an agentic development platform.

Skills

Python proficiency

ML system design

Robotics familiarity

Software engineering

Tools

PyTorch

JAX

Job description

Seed-stage infrastructure company in New York building an agentic development platform for the next generation of autonomous physical systems. *Remote opportunity available for SF Bay Area candidates while they establish a new office location.

The company's AI-native platform helps engineering teams model, validate, test, and continuously improve complex real-world systems in software before deployment. Rather than replacing simulation engines, it orchestrates the engineering workflow surrounding simulation, validation, testing, and continuous improvement, dramatically shortening engineering feedback loops while reducing dependence on expensive real-world testing.

A core part of this challenge is learning from both simulated and real-world data: building models that approximate complex physical systems, identify where simulation diverges from reality, improve predictions as new data arrives, and help engineers make better decisions about what to test next.

The founding team combines deep expertise across autonomy, AI, distributed systems, and enterprise software, and is assembling an exceptionally technical, in-person engineering team in New York City.

Summary

This isn't a traditional machine learning role.

You’ll build ML systems that help engineers understand, predict, and improve complex physical systems.

You’ll work at the intersection of machine learning, simulation, sensing, system identification, validation, and developer infrastructure. Some problems may involve learning surrogate models for expensive simulations. Others may involve predicting real-world system behavior, detecting discrepancies between simulation and reality, optimizing system parameters, or determining which experiments and simulations will provide the most useful information.

The feedback loop between simulation and reality is central to the work. Models need to improve as new simulation results, sensor measurements, experiments, and customer data become available.

You’ll also work directly with engineers and customers to understand unfamiliar physical systems, determine what should be modeled or learned, and turn those solutions into reusable capabilities for future customers.

The problems are rarely clean. You’ll encounter noisy sensor data, sparse observations, distribution shifts, incomplete physical models, unfamiliar hardware, and situations where ground truth is expensive to obtain.

Success comes from combining strong ML judgment with software engineering rigor and the ability to reason about real-world systems from first principles.

If you enjoy building ML systems where model quality has to survive contact with the physical world, this is one of the most interesting opportunities we’ve seen in Physical AI.

What You’ll Build

• Build ML models that learn from simulation, experimental, sensor, and real-world system data.
• Develop surrogate and learned models that approximate expensive or complex physical processes and enable faster engineering iteration.
• Build systems for system identification, parameter estimation, calibration, and optimization of physical systems.
• Develop methods for measuring and reducing the gap between simulated and observed real-world behavior.
• Build ML-driven workflows for anomaly detection, failure analysis, validation, and continuous improvement.
• Develop data and training pipelines that combine simulated and real-world observations.
• Design evaluation frameworks for understanding model accuracy, uncertainty, robustness, and generalization across operating conditions.
• Explore techniques such as active learning, uncertainty estimation, optimization, and adaptive experimentation to determine which simulations or real-world tests should be run next.
• Productionize models as reliable components of a broader engineering platform rather than isolated research prototypes.
• Work directly with customers to understand unfamiliar physical systems, then generalize those learnings into reusable ML and platform capabilities.
• Help define the ML architecture and technical direction of an agentic development platform for autonomous physical systems.

You’ll work closely with the founders and early engineering team across machine learning, platform engineering, simulation, AI infrastructure, and customer deployments.

What We’re Looking For

You likely have experience with several of the following:

• Building and deploying production machine learning systems using Python and modern ML frameworks such as PyTorch or JAX.
• Developing models from complex numerical, temporal, spatial, sensor, or scientific data.
• Working on problems involving robotics, autonomy, simulation, scientific ML, control systems, computer vision, industrial systems, or other domains where software interacts with the physical world.
• Building surrogate models, learned dynamics models, system identification methods, forecasting systems, or other models of complex system behavior.
• Designing rigorous evaluation approaches where model performance cannot be reduced to a single offline benchmark.
• Working with noisy, sparse, imperfect, or expensive-to-collect real-world data.
• Building end-to-end ML systems spanning data, experimentation, training, evaluation, deployment, and monitoring.
• Shipping production-quality ML systems rather than stopping at research prototypes.
• Operating effectively in ambiguous, fast-moving startup environments.

Experience with simulation, digital twins, robotics, physics-informed ML, reinforcement learning, optimization, active learning, or scientific computing is valuable, but we’re ultimately looking for exceptional machine learning engineers who enjoy reasoning about complex real-world systems.

You do not need to be a simulation expert. You should, however, be excited about applying machine learning to systems where physics, software, data, and the real world interact.

Prior startup experience is a strong plus.

Our ideal candidate combines strong ML fundamentals with excellent software engineering instincts, has high agency, cares deeply about model and system quality, and wants to own difficult technical problems from first principles through production.

About Us

Greylock is an early-stage investor in hundreds of remarkable companies including Airbnb, LinkedIn, Dropbox, Workday, Cloudera, Facebook, Instagram, Roblox, Coinbase, Palo Alto Networks, and many others. Learn more at https://greylock.com/

How We Work

We are full-time, salaried employees of Greylock and provide free candidate referrals and introductions to our active investments. This posting is for direct employment with one of our portfolio companies. We review every application and reach out directly when we believe there’s a strong potential fit.