About NovifyAI
NovifyAI builds focused AI products that turn complex technology into clear, useful tools. We work across models, infrastructure, and product experiences with a strong bias toward quality, ownership, and real outcomes.
About the role
As ML Engineer (Training Infra), Foundational Models, you will advance the data, training systems, and research behind capable foundational models. You will own meaningful work from problem definition through delivery and improvement.
The role combines independent judgement with close collaboration across NovifyAI. We value people who can understand the details, communicate the tradeoffs, and keep the work tied to a useful outcome.
What you'll do
- Design rigorous experiments and production-quality model development workflows.
- Build reliable systems for data, training, evaluation, and analysis.
- Work across research and engineering to turn promising ideas into measurable progress.
- Document assumptions, results, and tradeoffs so decisions remain reproducible.
- Raise model quality while improving efficiency, safety, and operational reliability.
What we're looking for
- Strong ownership: you can take a meaningful problem from an unclear beginning to a high-quality result.
- Clear written and verbal communication across technical and non-technical teams.
- Good judgement about priorities, tradeoffs, detail, and when to move quickly.
- Comfort working in a fast-moving environment where the process is still being built.
- Genuine curiosity about AI products and a willingness to learn the domain deeply.
Bonus points
- Experience shipping or supporting AI products in production.
- Evidence of unusually strong craft in your area of work.
- Experience collaborating across product, engineering, research, and go-to-market teams.
Why NovifyAI?
Join a small, ambitious team working across AI research, engineering, design, and go-to-market. You will have meaningful ownership from day one, work closely with people who care about their craft, and help shape products that are built to be used—not simply demonstrated.
We care more about your ability to own the outcomes described here than matching every line. If the work excites you and you believe you can do it well, we want to hear from you.