About the Role:
We're looking for a Data Scientist to join our Data Sciences and ML Engineering team. You'll be building, shipping, and improving the models and LLM-powered systems that sit at the core of our security products — the pieces that make the difference between a tool that flags noise and one that helps defenders find what matters.
This is a hands-on role on a small, focused team. You'll have real ownership over the models and pipelines you build, close collaboration with engineering and product, and the runway to go deep on the hard problems.
What You’ll Do:
Security is a domain where the adversary is adaptive, the signal is rare, and the cost of a miss is real. That makes for interesting modeling problems — ones where off-the-shelf approaches rarely carry you the whole way, and where careful research, solid experimentation, and production rigor each matter for our success.
Your work will span a few areas:
- Model development and research. Building classifiers, detectors, and scoring models on messy, high-stakes security data. Designing experiments, evaluating trade-offs, and iterating on architectures — not just hyperparameters.
- LLM agent systems. Shaping the prompts, context, tool-use patterns, and supporting content that drive our LLM agents.
- Production delivery. Shipping models behind real traffic, monitoring them, and improving them over time.
- Evaluation and iteration. Building the evaluation harnesses and feedback loops that let us know whether a change is actually an improvement — often the hardest part of the work. Our models only improve for customers when our evaluations highlight what really matters.
Who You Are:
Production experience is the single most important thing. We'd like to see around 3–4+ years of experience delivering models into production environments where they've had to perform, be maintained, and evolve. That's the background that tends to set people up for success here.
Beyond that:
- Depth in ML fundamentals. You understand model architectures and can reason about why a given approach is or isn't a good fit for a problem. You've moved well past treating models as black boxes and past tuning that stops at sample weights and decision thresholds.
- Willingness to experiment. You're comfortable trying genuinely novel approaches when the standard playbook runs out, and you can tell the difference between a promising result and a fragile one.
- Strong engineering instincts. Your code is something teammates can read, extend, and trust in production. You think about reproducibility, testing, and handoff — not just whether something runs on your laptop.
- Experience with LLMs in practice. You've worked with LLM-based systems in some real capacity — prompting, context design, tool use, evaluation, or fine-tuning — and have opinions shaped by actually shipping things.
- Comfort with ambiguity. Security problems rarely come with clean labels or clean data. You're able to frame problems, scope them, and make progress without a fully paved path. You’ll help highlight ambiguity and reason about how to make progress even when humans don’t all agree on one single answer.
- An advanced degree (MS or PhD) in a technical discipline. This doesn't have to be in data science or ML specifically — strong backgrounds in CS, statistics, physics, math, engineering, and related fields are all welcome. Your on-the-job experience is what matters the most.
We want to be upfront about research and publications: while we're supportive of engagement with the broader research community, our team's focus is firmly on shipping. Publishing papers and attending conferences can absolutely happen, but they aren't the center of gravity of the role. If your primary goal is academic output, this probably isn't the best fit — and we'd rather say that clearly than have it be a surprise.