Senior Machine Learning Engineer

Sumsub is recognized as a Great Place To Work® in the US and rated 4.3 out of 5 on Glassdoor. Our 1,000+ people around the world back it up, too: 92% are proud to be Sumsubers, and 90% say their work is interesting and challenging.

Now we are looking for an experienced Machine Learning Engineer to build, deploy, and own end-to-end ML solutions that combat sophisticated fraud, deepfakes, and identity theft. This role is an exciting opportunity to have true product autonomy, see your models protect millions of users in real time, and scale a platform that handles millions of verifications a day.

• Remote-first, trust-based culture. Work from the place that works best for you. No mandatory office days, no attendance trackers. In some locations, we provide offices or coworking spaces, but the choice is yours.
• True flexibility. We do not fix you to a 9-to-5 schedule. You can adjust your working hours when needed, as long as your day stays productive and in sync with the team.
• Extra time off. Your birthday is a holiday here. Add to that 10 personal days each year, seven sick days without paperwork, and extra time to enjoy Christmas and New Year. Time to rest is part of the deal.
• Work that matters. Our mission is to build a digital world that is secure, accessible and inclusive for everyone. From fighting fraud to making online services easier and safer to use, your work will have a real impact on how people experience trust online.
• Compensation. We offer fair and transparent pay, benchmarked to the market.
• Truly global. We work across continents and time zones, with teammates and customers from all over the world. You will run campaigns that cross borders, cultures, and languages, and see your ideas land worldwide.
• Growth built in. Clear goals, open feedback and personal development plans. We support your progress with learning opportunities and by covering role-specific events, from design conferences to marketing forums.
• Team offsites. Sometimes just Slack is not enough. That is why we meet in person a few times a year. Trips are fully covered, so you can meet, collaborate, and recharge together.
• Getting you set up. We make sure you have access to the tools and hardware you need to do your work well.
• Friendly by design. Our logo is a dog for a reason. We keep things human, open and kind. We welcome individuality, quirks and different perspectives, because that is what makes our work smarter and more fun.

• 4+ years of experience as an ML / CV Engineer in fast-paced, high-load product environments.
• Strong background in deep learning frameworks (PyTorch is our go-to) and solid production-grade engineering skills.
• Fluent with the data side of the work: labeling, collection, and navigating the data space to decide what to include, what to leave out, and how those choices affect training. Solid grasp of how the training distribution relates to what the model sees in production — sampling, skew, and shift, and how to catch it before it bites.
• Familiar with current research, the latest baselines, and modern architectures.
• An autonomous product owner: capable of taking an idea from inception to implementation under the pressure of a competitive market.
• A proactive communicator: you easily align with PMs and backend teams, and you don't go quiet if something breaks - you find the solution.

Nice to have:

• Experience standing up model serving and the infra around it (Triton or FastAPI, queues, autoscaling, monitoring).
• Comfortable designing and reading online experiments / A-B tests.
• Experience with big data and stream processing tools (Kafka, Flink, Trino, or ClickHouse).

• Own the full ML lifecycle: from data collection, labeling, and training through optimization, production rollout, and ongoing quality assessment
• Train CV/ML models to detect deepfakes, segment documents, and uncover complex anomaly patterns in real time.
• Balance research and speed: know when to build a full deep learning model and when to ship a simple, clever heuristic that solves the problem fast.
• Serve models efficiently with FastAPI and PyTorch, integrating cleanly with our high-load Java backend.
• Build and monitor ML pipelines and dashboards with Airflow, ClickHouse, and Superset so models stay reliable under heavy load.