Applied and Product ML
In one line
You turn models into things people actually use — fine-tuning, evaluation for product use cases, data pipelines, shipping features. The bridge between “the model works in a notebook” and “the model works for a million users.”
What it actually is
This is the most “industry job” of the tracks, and the one with the most roles overall. Applied / ML engineers sit between research engineering and product: they take existing models and make them solve real problems — fine-tuning for a domain, building eval harnesses for a specific feature, wrangling data pipelines for production, making latency and cost acceptable. The interview loop looks like research engineering with more product sense layered on.
Fewer programs brand themselves as “applied ML fellowships” specifically — this track is more often entered through residencies, internships, and new-grad roles than through the safety-flavored fellowship pipeline. But it’s a legitimate and often underrated path, and it converts: applied → research engineering is a well-worn road.
What you actually do day to day
- Take a model and make it good at a specific real task (fine-tune, prompt, evaluate, iterate).
- Build the eval that tells you whether your change actually helped users.
- Own a feature end to end — data, model, serving, monitoring.
- Trade off accuracy vs. latency vs. cost vs. ship date in the real world.
What you have to do to get in
The path
Strongly portfolio-driven. The best evidence is something that works end to end — a small app or tool that uses an ML model, deployed, with you able to explain every decision. Routes in: the OpenAI Residency (takes strong builders from adjacent fields), lab/company residencies and internships, and new-grad ML roles. If either fellow is currently enrolled, a internship is an excellent applied on-ramp.
Skills required
Mostly overlaps with Research Engineer (see Technical Skills), weighted toward:
- Strong software engineering, including the boring production parts — testing, deployment, monitoring.
- Practical ML: fine-tuning, prompting, evaluation, RAG, working with APIs and open models.
- Product sense: what does the user actually need, and what’s the simplest thing that delivers it.
- Less heavy math and original-research skill than the research tracks.
Is this you?
Signs you lean applied/product
- You like seeing real people use the thing you built.
- You care about the whole pipeline, not just the clever model part.
- You’re pragmatic — “good and shipped” over “perfect and theoretical.”
- You want a clear, fast route to working in the field.
Pointers & extra resources
- The OpenAI Residency page is the model description for the “talented builder from an adjacent field” route.
- Build something small and real with an open model or an API this month — it’s the best possible portfolio piece for this track. (The portfolio mindset)
- More in Reading and Courses.