СШАSoftware Engineer, Machine Learning Systems
2 нед. назад
СШАMiddleOnsite
machine learningpythondeep learningai toolsdata processing
Software engineer to maintain machine learning system infrastructure and develop tools for atmospheric data processing and model inference at .
Обязанности
- Keep forecasts running. Build the tooling that runs our weather models and detects failures. If a cloud node goes down, our systems need to recover and keep producing forecasts.
- Process atmospheric data in real time. Build pipelines for incoming observations, account for late or missing data, and make it clear when something needs attention.
- Show us what’s happening. Build tools to visualize data processing and model inference as they run. The team should be able to see how a run is progressing, where it’s slowing down, and what went wrong if it fails.
Будет плюсом
- Experience with data pipelines, distributed systems, or infrastructure.
- Experience with scientific computing, weather data, or geospatial systems.
- Familiarity with ML inference, GPU computing, or PyTorch (again, not required).
Условия
- Base salary: $120,000–$200,000, depending on experience and scope.
- Location: In person at 1600 Bridge Parkway, Redwood Shores, CA.
- Benefits: 401(k), health, dental and vision insurance, unlimited PTO, stock options, and office food and beverages.
Другое
- designs, manufactures, and operates Atlas, a global constellation of smart weather balloons. The atmospheric observations they collect help drive WeatherMesh, our AI weather model .
- Our model, WeatherMesh, runs 24/7, producing global forecasts every hour. Our deep learning research team is small and iterates quickly, but sometimes R&D gets ahead of the internal tools that support it. We’re looking for a talented and curious software engineer to keep model inference running reliably, process atmospheric data as it arrives, and build tools that show us what’s happening in both.
- In this role, you’ll work with our deep learning team and backend and frontend engineers. We have problems for you to tackle right away, and we also want someone with ideas about what to build next. You’ll help figure out what’s needed, build it, and be responsible for how it works once it’s running.
- Prior machine learning experience is not required! You can come in with no ML background at all and be successful. Where you lack understanding, you’ll dig into model code, ask researchers questions, and use AI tools to get up to speed. Essentially, learn what you need to build and go build it. We care much more about how well you learn and your approach to problem solving than whether you have RAG or LLM fine-tuning on your resume. [Note: We’re not trying to shade ML engineers - we have many great ones on our team. Yes, those skills can be useful in other domains, but they’re not prerequisites for this role. We’re bringing them up to emphasize that, without good software taste and a knack for learning, ML skills alone are not what we're looking for.]
- You’ve owned software in production. You’ve been responsible for a system after launch: how it performs, how it fails, and how it recovers. That experience has given you a feel for the monitoring, reliability, and operational discipline a critical system needs.
- You’ve built substantial software. You can walk us through a hard technical problem you solved, why you chose your approach, and what you learned. You’re comfortable reading unfamiliar code and tracking down bugs that involve more than one part of a system. Bonus points for good UI design and a portfolio we can see.
- You think beyond the immediate fix. You can solve today’s problem while recognizing what should become reusable infrastructure for future work. You have ideas about how to improve the systems you own and want to help decide where we go next.
- You’re curious and motivated to build. You ask questions, investigate how things work, and turn what you learn into something useful. You don’t sit idly while waiting for guidance; you can’t help but poke around.
- You use AI and agentic tools like a bandit. You use them everywhere you can: to explore code, learn about unfamiliar systems, test ideas, and build quickly. You understand the code you ship and can debug it when it doesn’t work. Most importantly, you know these tools are powerful, but you can’t outsource thinking to them entirely. After all, you’re still on the hook for what gets built.
- You want to work on a dynamic team and see your work put to use quickly.
- We’re flexible on programming languages, frameworks, and years of experience. Show us what you’ve built and how you think through problems. You can learn the stack here, and honestly, things change quickly.