What we're building

AI agents are showing up in enterprise infrastructure faster than platform teams can identify them and security teams can govern them. The technical surface is new - agents make decisions, call tools, hold credentials, and reach across systems in ways traditional models weren't designed for. The companies running thousands of these agents in production have problems nobody has clean answers to yet. Tigera has built a new product to tackle this space. We're a focused engineering team working on the hard parts: detecting agents at runtime, understanding their behaviour, distinguishing legitimate activity from misbehaviour, and giving security teams the controls they need without slowing the platform down. The problems span machine learning, distributed systems, applied AI, and security research. We're early enough that the work you do will shape the product, and far enough along that you'll see real customers using what you build.

You will own the machine learning and applied AI side of this product, turning the considerable agent telemetry our platform captures into the detections, risk scores, and behavioural baselines that decide whether an AI agent gets to run inside a regulated enterprise. The modelling work is central to what makes the product work, and you'll own it end to end. Including classification from runtime telemetry, behavioural threat detection and using LLMs to bridge the gap between security intent and machine-enforceable policy. You will also be the AI/ML voice in our broader architecture decisions - the telemetry pipeline, product features and roadmap, model deployment and versioning, A/B testing of detection models in production. Our data infrastructure is in place; the ML systems built on top of it are yours to design.

What you'll bring

Required:

• 5+ years of professional ML engineering experience, with at least two years building and deploying production ML systems
• Strong fundamentals in classical machine learning - gradient-boosted trees, regression, classification, evaluation methodology, feature engineering, dealing with class imbalance and noisy labels
• Experience with anomaly detection or time-series modelling in adjacent domains (fraud detection, observability, recommendation systems, fault detection etc)
• Hands-on experience using LLMs for applied tasks beyond chatbots - function calling, retrieval-augmented generation, prompt engineering, fine-tuning, evaluation
• Python and the standard ML ecosystem (scikit-learn, PyTorch or TensorFlow, pandas)
• Comfort working with large-scale telemetry data - ClickHouse, BigQuery, Snowflake, Spark, or equivalent
• Strong communication skills, including excellent writing skills

Nice to have:

• Prior experience in security, infrastructure, or systems-adjacent ML
• Familiarity with eBPF, kernel telemetry, or low-level systems observability
• Experience deploying ML models in latency-sensitive paths (sub-millisecond inference)
• Open-source contributions to ML tooling or applied AI projects
• Experience with model versioning and various frameworks (e.g. MLflow, Weights & Biases, BentoML, or equivalent)
• Background in interpretable ML making model decisions defensible to enterprise customers and auditors

How we work

Small team, high autonomy. You will own significant chunks of the product end-to-end rather than working through layers of management. We plan and ship fast and review architecture decisions openly. The CTO is the engineering leader and the team is flat below that. You will report directly to the CTO.

We use Claude Code internally and our engineers move quickly with AI assistance, we expect you to as well, and to bring ideas about where AI tooling makes the team more effective.

Cork-based. In-office presence twice a week every Monday and Wednesday.

External visibility can also be part of the role. You will have the opportunity to potentially publish your research, speak at conferences (KubeCon, AI Engineer Summit, security research venues), and represent Tigera in the technical community. This is a real opportunity to build a profile in a fast-growing space.