AI Research Engineer

Remote: yes Posted: 2026-08-03

ABOUT COGENT

Cogent is an Applied AI Lab building the next generation of AI agents for cybersecurity. AI has fundamentally changed how attacks happen, allowing malicious actors to operate at unprecedented speed and scale. Cogent’s "AI Taskforce" assesses petabytes of enterprise data to remediate these issues before critical breaches occur.

To stay at the cutting edge, we blend frontier research with real-world execution. Alongside our core product work, Cogent Research serves as our applied AI lab, providing the research horsepower needed to make truly agentic security workflows a reality.

Since coming out of stealth, Cogent has experienced rapid growth. We partner with Fortune 500 companies to secure some of the most complex production environments in the world.

We’re backed by Greylock and we’ve built a team with the best minds in applied AI. Our team is comprised of people from:

- Top universities like Stanford, Berkeley, Penn, Duke, Carnegie Mellon, Waterloo

- Unicorn, high-growth companies like Scale AI, Databricks, Stripe, Tesla, Coinbase

- World class cybersecurity experts from Wiz, Abnormal AI, Zscaler

- Preeminent research labs like Deepmind and SAIL

Cogent Research is the applied AI lab at the heart of Cogent's mission. As an AI Research Engineer, you will work at the frontier of automated reasoning and AI-driven code analysis — inventing novel solutions, prototyping fast, and pivoting when first ideas don't pan out.

This is not a role for those who need a playbook. We are looking for someone who thrives in ambiguity, drives work proactively and independently, and is deeply aligned with Cogent's mission to make the world's infrastructure safer.

- Conduct original research at the intersection of AI and automated code reasoning, advancing the state of the art in areas such as static/dynamic analysis, formal verification, and automated testing

- Invent and apply novel AI techniques to challenges in code reasoning and code generation, from early hypothesis through working prototype