<aside> 👉

Apply for Senior Applied AI Engineer

</aside>

💙 About VIDOC:

We are a team of hackers building an AI Security Engineer: security automation that aims to match human-level performance in security code review. We are developing state-of-the-art AI Agents for cybersecurity and integrating them into autonomous systems. We're a 13-person team split between Poland (mainly Gdańsk) and San Francisco. We've raised $6M so far and are raising more. We're part of the NATO DIANA program and actively engaged in other NATO-related initiatives. Our clients operate in industries where cybersecurity is a crucial element of business.

About the Role:

As an Applied AI Engineer, you will build, improve, and scale the agents behind Vidoc's AI Security Engineer. You will spend most of your time improving agent performance through engineering, evals, datasets, and experimentation.

We are looking for a first-principles thinker and an excellent software engineer. You understand that agent quality depends on the complete system around the model. This includes the harness, workflows, context, tools, data, and evaluation methods.

You will own agent capabilities from initial research through production. You will diagnose failures, design experiments, build evaluation systems, and monitor agent behavior after release.

The work does not stop at the agent boundary. You will integrate agents across our backend, product, and infrastructure. You may build an API, change a user workflow, or improve production infrastructure when the agent requires it.

You will work with proprietary and open-weight models across cloud and on-prem environments. You will balance model quality, cost, latency, privacy, security, and deployment constraints.

Your previous role may have been in ML engineering, data science, AI research, or software engineering. We care more about what you can build, how you reason about models, and whether you can take complex agent systems into production.

What you'll do

• Improve agent performance through harness, workflow, and context engineering. Work on tool use, retrieval, memory, prompts, state, recovery, and model routing.
• Build evals and curate datasets from production failures, customer feedback, security expert input, and synthetic examples. Use them to measure improvements and prevent regressions.
• Design complex agentic systems for security workflows, including planning, tool execution, human review, safety, recovery, and traceability.
• Evaluate proprietary and open-weight models across cloud and on-prem environments. Determine when model serving, fine-tuning, or reinforcement learning provides measurable value beyond system-level improvements.
• Work directly with founders, engineers, security experts, and customers to turn security problems and production failures into measurable agent capabilities.
• Own agent capabilities end to end across backend services, product workflows, infrastructure, deployment, observability, latency, cost, and reliability.

Who you are

• You are an excellent software engineer who can design, implement, test, debug, and operate complex systems.
• You have built serious LLM or agent systems beyond basic API wrappers and demos. You understand that agent quality depends on the complete system around the model. You can determine whether a problem belongs in the harness, workflow, context, tools, data, or model.