Machine Learning Engineer Brooklyn, NY On-site $190,000 – $245,000 · per year · USD

Nodi; New York (NY); On-site; USD 190,000 - 245,000; Full time

Benefits offered by this job

100% employer-covered medical, dental,

unlimited PTO

14+ company holidays

2 weeks of company shutdown at EOY

401(k)

mental health support

Job summary

Radical AI in Brooklyn Navy Yard is seeking a generalist Machine Learning Engineer to turn ML research into production-ready capabilities inside our platform that runs our autonomous lab. A model that can't be trained reliably, served efficiently, and trusted by the scientists using it doesn't count.

Your job is to make sure it does. This hands-on role spans the full ML lifecycle, from debugging training runs to deploying services, with a focus on reliability, latency, and cost while maintaining

Qualifications

• 4+ years in ML engineering, research engineering, or a closely related software role.
• Strong Python and software engineering discipline: testing, packaging, and maintainable design.
• Hands-on experience with PyTorch: training, evaluation, and debugging models.
• Experience building data pipelines on cloud infrastructure with GPUs.

Responsibilities

• Turn research ideas into working implementations and surface constraints early.
• Build training pipelines, datasets, configurations, failure triage, and compute efficiency.
• Make experiments reproducible by tracking code, data, environments, and artifacts.
• Evaluate models for product use with data-driven decisions on readiness to ship.
• Deploy and integrate models with services, ensuring observability and cost awareness.
• Improve inference performance and partner with product and platform teams.

Skills

Python

PyTorch

ML production

Training pipelines

Cloud data pipelines

Tools

Containers

GPU workloads

Experiment tracking

Model deployment

Job description

Compensation $190,000 – $245,000 · per year · USD

The Role

We're looking for a generalist Machine Learning Engineer to turn ML research into capabilities that hold up in production, inside the platform that runs our autonomous lab. A model that can't be trained reliably, served efficiently, and trusted by the scientists using it doesn't count. Your job is to make sure it does.

You'll work side by side with our ML Researchers: designing experiments together, building training pipelines, evaluating models, and taking what works into production. You'll own the engineering that makes that possible, from data preparation and reproducible training to inference performance, product integration, and monitoring after release. Nobody hands you a spec; you'll help decide what's worth building, then build it.

This is a hands-on role across the full ML lifecycle. One week you're in the weeds with a researcher, debugging a training run or implementing a new approach. The next you're standing up a serving endpoint, running down a production failure, or rebuilding the pipeline that gets the next model into users' hands.

This is not a research role, and it's not a pure infrastructure role. We need strong software engineering fundamentals, real experience training and deploying models, and the judgment to trade model quality against reliability, latency, and cost, and to defend those calls when researchers push back. If your instinct is to hand off a model rather than own how it performs, or to wait for clear requirements before moving, this isn't the right fit. Success is simple: research improvements become measurable improvements in Radical's product.

What You'll Work On

• Partner with ML Researchers. Turn research ideas into working implementations, co-design experiments, and surface engineering constraints before they become blockers.
• Build and run training pipelines. Own the workflows for training and fine-tuning models: datasets, configurations, failure triage, and compute efficiency.
• Make experiments reproducible. Track code, data, environments, and model artifacts so any result can be reproduced and built on with confidence.
• Evaluate models for product use. Build the datasets and tests that measure model quality, expose failure modes, and decide, with evidence, whether a candidate is ready to ship.
• Deploy and integrate models. Turn research code into tested, maintainable components and services, and work with product and platform engineers to put them into real scientific workflows.
• Improve inference performance. Profile and optimize latency, throughput, memory, and cost without giving up the quality the product depends on.
• Own performance after launch. Monitor model and service behavior, run down regressions, keep releases and rollbacks safe, and feed what you learn into the next round of research.
• Improve how the team ships ML. Build the shared tooling and automation that shrink the path from a promising experiment to a dependable product capability.

What We're Looking For

• You've taken ML models from experimentation and training all the way into production, and owned the systems that kept them running.
• 4+ years in ML engineering, research engineering, or a closely related software engineering role.
• Strong Python and real software engineering discipline: testing, code review, packaging, and system design that someone else can maintain.
• Deep hands-on experience with PyTorch or a comparable framework. You've implemented, trained, evaluated, and debugged models yourself, not just called them.
• A working command of dataset quality, train/validation splits, overfitting, data leakage, and why offline metrics so often lie about production performance.
• Experience building data pipelines on cloud infrastructure, with containers and GPU workloads.
• Experience deploying models through services or batch pipelines, with real attention to observability, reliability, and operating cost.
• You engage with researchers as a peer, not through a translator. You pick up unfamiliar approaches quickly and make sound implementation calls while the requirements are still moving.
• A generalist's range and an owner's instinct. You move between model code, data, infrastructure, and product integration, and you don't stop until the whole system works.
• Able to work on-site at the Brooklyn Navy Yard.

Nice to Have

• Experience with LLM training or fine-tuning, model serving, or agent systems.
• Distributed training, GPU profiling, or inference optimization experience.
• Familiarity with experiment tracking, dataset and model versioning, and automated model release workflows.
• Experience with scientific ML, optimization, reinforcement learning, or sequential decision-making.
• A background in scientific computing, materials, chemistry, robotics, or lab automation.
• Experience shipping ML on a small team where the roadmap changes as fast as the science.

What Success Looks Like

In your first six to twelve months:

• Models built with our researchers are live in Radical's product and delivering measurable improvements.
• Training and evaluation workflows are reproducible and reliable, and the rest of the team actually uses them.
• Every model release has clear quality criteria, automated checks, production monitoring, and a rollback path that works.
• The path from promising experiment to production release is measurably faster and far less manual.
• Production failures and user feedback flow directly into new datasets, evaluations, and research priorities.
• Model quality, inference performance, and operating cost are understood, tracked, and improved together, not traded off by accident.

Work Arrangement

This role is fully in person at our lab in the Brooklyn Navy Yard. We're an in-person team.

Compensation and Benefits

Compensation varies based on experience and scope. We offer 100% employer-covered medical, dental, and vision insurance, unlimited PTO, 14+ company holidays, 2 weeks of company shutdown at EOY, 401(k), and mental health support.

Radical AI is an equal opportunity employer. We do not discriminate on the basis of race, color, ancestry, national origin, religion, sex, age, sexual orientation, gender identity and expression, marital status, disability, or veteran status.