Engineering
Senior ML Engineer
What we do
Every company is racing to put AI into everything - and very few of them can answer the questions that come right after: what is our AI actually allowed to do, what did it do last week, and can we prove it? We're building the layer that answers those questions, so teams can move fast without losing control of what their AI is doing.
Our customers are the people who carry the weight when AI goes wrong: the banks, hospitals, and insurers who have to sit across from a regulator, a board, or a customer and explain exactly what happened. We give them the confidence to say yes to AI without flying blind. It's early, the problems are complex and consequential, and the work matters - which is exactly why it's worth doing.
About this role
As a Senior ML Engineer, you'll own the models at the heart of the product - the ones that read what an AI system is trying to do and decide whether to allow it, change it, or block it. They catch a prompt trying to jailbreak the system, spot sensitive data before it leaks out, and tell the difference between a routine action and a genuinely risky one.
You'll own that work from the first question all the way to production: defining the real problem, working through complex and imperfect real-world data, researching and prototyping approaches, and validating them until the results are ones you can trust. You'll make the decisions that matter - technical and product alike - about where a model should be confident, where it should defer to a human, and what "good enough to ship" means when a wrong answer carries real consequences.
It's demanding work: the models have to be both accurate and fast, because they run on live customer traffic where every mistake is visible and costly. If owning a hard, open-ended problem end to end - from the first ambiguous question to a system running reliably in production - is the kind of work you're drawn to, you'll feel at home here.
What you'll do
- Own our detection models end to end, from data and training through evaluation and production.
- Make them fast: a safety model that slows every request to a crawl helps no one.
- Build the evaluations and red-team sets that keep quality honest as the threats keep changing.
- Work across engineering and product so your models plug cleanly into the wider system.
- Keep your models healthy in production - watch for drift, catch regressions, and roll out changes safely.
- Partner with product owners and stakeholders, turning hard ML trade-offs into clear, confident decisions.
- Raise the bar for how we do ML - mentor teammates, review each other's work, and share what you learn.
- Stay close to the research and bring the best of it into the product.
What we're looking for
- A Master's or PhD in Computer Science, Data Science, Mathematics, or a related field - or a Bachelor's with 8+ years of relevant experience.
- 5+ years building and shipping ML systems in production, with real ownership of models that served live traffic.
- A strong applied background in NLP or LLM safety - classification, sequence modeling, or content moderation at scale.
- Expert in Python and the modern ML stack - PyTorch, Hugging Face Transformers, scikit-learn, spaCy - and comfortable under the hood, not just calling the APIs.
- Comfortable with large-scale data tooling (Spark / PySpark, Ray, Pandas, Apache Arrow, Kafka) to build the pipelines that feed training and inference.
- Experience serving models in production at low latency - vLLM, ONNX Runtime, or NVIDIA Triton, plus quantization and distillation to hit tight latency budgets.
- Fluent in the production stack (Docker, Kubernetes, CI/CD, AWS or GCP) and the stores behind ML systems (PostgreSQL / pgvector, Redis, S3, Snowflake).
- Familiar with experiment tracking and orchestration (Weights & Biases or MLflow, Airflow, DVC) and disciplined about reproducibility.
- A real feel for performance - you care about the slow tail, not just the average.
- Discipline about evaluation - you don't trust a model you can't measure.
- Excellent communication skills - you can take an abstract, complex problem and clearly explain both it and your path to a solution for product owners and non-technical stakeholders.
- Comfortable owning open-ended problems in a fast-moving, early-stage team.
Nice to have
- Experience with adversarial ML, prompt-injection defense, or abuse detection.
- Exposure to AutoML tooling (Optuna, Ray Tune, AutoGluon, or similar) for automated hyperparameter search and model selection.
- Familiarity with regulated worlds (finance, healthcare, government) and the bar they set for evidence.
- Lower-level systems chops - Rust, C++, or CUDA - for the moments where milliseconds matter.
This role is for you if you are
- Passionate about working at the cutting edge, where cybersecurity, data, and AI all come together on problems that actually matter.
- A problem solver who loves teaming up to crack things no one has cracked yet.
- At your best on a small, global, fast-moving team that's sharp, committed, and genuinely fun to be part of. :)
Not quite your role? Browse all open roles or introduce yourself.
DataStrict is proud to be an equal-opportunity employer. We celebrate difference and are committed to building an inclusive team where everyone can do their best work, regardless of race, color, religion, sex, sexual orientation, gender identity or expression, national origin, age, disability, veteran status, or any other characteristic protected by law. If you need a reasonable accommodation at any point in the process, tell us at [email protected] and we'll make it happen.
By submitting an application, you agree to DataStrict's Terms of Service and Privacy Policy.