Cracken is a fast-growing Silicon Valley-based startup built by elite nation-state and commercial operators who defended critical cyber infrastructure during the war in Ukraine, researched AI and cybersecurity at MIT and Kyiv Polytechnic, and led teams at Apple, Google, Palo Alto Networks, HackerOne, DIU, Comcast, HP, and more.
We tame Cracken, an AI Agent for Cybersecurity with human-in-the-loop. Our flagship product, is an agentic AI copilot drastically scaling cyber assessments for enterprises and governments.
We are growing and seeking an AI/ML Engineer to help us pioneer the future of cybersecurity.
ResponsibilitiesDesign and implement LLM-based pipelines and agentic systems for secure-code understanding/generation in production SaaS.
Setup and manage infrastructure for training, evaluating, and serving LLMs (MLOps responsibilities).
Collaborate with data-science, product and security teams to embed ML into customer features.
Conduct ML/AI research and prototyping to explore and validate new techniques and tools.
Optimise existing models for latency, accuracy and cost; introduce new techniques as the field evolves.
Safeguard data integrity and security across the entire ML lifecycle.
3+ years leading applied-AI projects with evidence of scaling LLM or multi-agent systems to production.
Deep experience in Python.
Experience with AI agent frameworks:
— implementation of Generative AI Agents using Langchain / Haystack and others;
— implementation of RAG using LlamaIndex and other Advanced RAG frameworks;
— evaluation of AI Agent Performance;
— integration of AI Agents into Existing Systems.
Experience with cloud technologies, specifically in the ML/AI context.
Experience with MLOps: infrastructure setup, model evaluation, serving pipelines.
Cybersecurity product background: worked on security-focused software or platforms.
Experience containerising & orchestrating ML services (Docker, k8s) and automating workflows.
Publications or notable open-source contributions in ML / security.
Background in code-analysis or application-security domains.
Real-World Impact: Validated in nation-state operations and supported by internal researches.
Team's Unique Motivation & Expertise: Ukrainian-heavy international team with Ph.D.s, professors, and top-tier industry veterans, driven by intrinsic passion forged through exposure to the hardships of war.
Impeccable Timing & Positioning: At the cutting edge of AI, cybersecurity, and autonomous systems, Cracken is best positioned to win the race against adversaries.
Top Skills
What We Do
Release the Cracken. Defend what matters.
Cracken is the first Red-AI Copilot built for real offense, not an ordinary CNAPP or endpoint security. We arm defenders with nation-state grade attack logic at machine speed, under full human command. Not black boxes. No fluff.
Born on the hardest frontlines of cyber conflict -- Ukraine, US, EU -- Cracken is built by operators who know what it means to face adversaries that don't play fair. We are turning that battlefield DNA into the only agentic AI platform that simulates live, adaptative attacks across cloud, endpoint, and hybrid environments.
Why it matters:
⚡Human-in-the-Loop control: every action is transparent, auditable, and reversible. CISOs don't trust black-box AI. We give them speed and oversight.
⚡Accuracy over noise: no posture fluff, no false positives. Only validated, exploitable risk that matters.
⚡100% faster coverage than manual red teams.
⚡Undetected vulnerabilities caught before exploitation (proven in North America and Ukraine's critical infrastructure).
⚡Zero black-box risk with audit trails, rollbacks, and operator trust at the core.
⚡Regulatory ready for CISA, EU AI Act, and global oversight.
⚡Quantifiable adoption with seamless integration into CI/CD multi-cloud and hybrid environments.
Cracken is not another CNAPP, CIEM or compliance dashboard. Those show posture. We show truth. Attackers don't wait, neither should you.
👉Investors, customers, and security leaders: join us in weaponizing defense. Let's release the Cracken!
www.cracken.ai
.png)
