Senior ML Manager, Runtime Defenses
Description
Securing generative AI systems and agents is one of the most important problems of our generation, and Check Point Zurich (formerly Lakera) is one of the key players in shaping the field. We crowdsource adversarial attacks through gamification, define benchmarks, and publish what we learn to help the industry. We turn that knowledge into enterprise-grade protection; our customers, ranging from fast-moving AI-native companies to regulated enterprises, face the same challenges: the upside of enabling AI for coding, deploying agents to production, mixing sensitive and untrusted data is not possible without the right security measures.
The Runtime Defenses team builds the ML systems behind that protection, end to end, from training data to runtime inference. We work closely with two full-time product managers who have deep ML backgrounds, and we stay close to customers. We experiment a lot, but always with a pragmatic focus on what improves the product. We regularly collaborate with an in-house research team.
What you will do
- Lead a group of 5-6 engineers within the Runtime Defenses team, align them on a clear mission and roadmap in collaboration with Product and senior leadership
- Start as a hands-on tech lead, or as tech lead and manager, depending on your background. You learn our systems from the engineers who built them
- Coach talented junior engineers and aspiring seniors toward larger scope
- Own product definition and the quality of training and inference, together with Product
- Work across the full system: training data collection, benchmarks and evals that define product behavior, complex training pipelines, and runtime inference on a chain of models where latency and accuracy both matter. We also own part of inference optimization
- Stay close to customers: follow how defenses perform in their evaluations and in production, and fix the errors that customers or our traffic-analysis agents surface
Who you are
- You have 5+ years of experience leading ML projects, ideally including managing people. You stay close enough to the technical work to review designs, experiments, and results yourself
- You have shown strong data and quality skills: building datasets, defining evals and benchmarks, and finding and fixing failure modes
- You enjoy coaching engineers and helping them grow into larger roles
- You are curious about product and customers, and you like working with them directly
- Experience with language models or LLMs is a plus. We also welcome strong candidates from other ML domains
- Interest in cyber security is welcome, but not required
Why you’ll like it here
- Talented team. Our engineers combine strong ML skills with low ego, a habit of helping each other, and real curiosity about product and customers
- Frontier technology. You work with state-of-the-art models, LLMs, and agents in a field that moves fast. The problems are hard and open, and side experiments can change the roadmap
- Learning from production. We own the product end to end, so we see how customers use it at the leading edge of AI adoption, and that shapes what we build next. New engineers ship to production quickly
Example projects you would be working on
- extending the text-based runtime protection with native multi-modal support;
- making decisions on the latency-quality tradeoffs, combining insights from data and customers;
- regularly revisiting model and training choices in the context of model architecture and innovation and hardware availability;
- supporting engineers working on new evals and reviewing existing ones.
Skills
Sign in to read and join the discussion.