Schedule a demo
All roles

AI Researcher, AI Safety & Security

Role description

You'll lead research on how AI systems behave, how they can be attacked, misused, or fail, and how to defend them. That spans everything from a single chatbot or RAG pipeline to multi-agent systems, third-party agent harnesses, and the frontier models underneath them. Tinycode runs autonomous red-teaming, evaluations, and remediations for enterprises deploying AI and for the labs building it, and your research will power everything our platform does.

A core part of the role is building custom AI evaluations, from custom frontier capability evals for labs' training and release gates to domain-specific assessments for enterprises in regulated industries. We bring in subject-matter experts for specific threat domains, and you own the science that turns their expertise into rigorous, repeatable evaluations. This is a builder's role for a researcher: you'll set the research bar, write code, run experiments, and ship research into product alongside the founding team.

You might be a fit if you have

  • Strong research experience in machine learning, especially with LLMs and agentic systems
  • Deep hands-on experience in one or more of: AI evaluations, adversarial ML or LLM red teaming, agent safety, alignment, or AI security
  • A track record of research output, such as peer-reviewed papers at top venues (NeurIPS, ICML, ICLR, ACL, EMNLP, USENIX Security, IEEE S&P, or similar), widely used open-source tools or evals, or equivalent work at an AI lab, security company, or safety institute
  • The ability to design novel research approaches for questions nobody has answered yet, and to adapt quickly as the questions change
  • Proficiency in Python and comfort building research infrastructure, not just running notebooks
  • Experience working in multidisciplinary teams, including with domain experts who aren't ML researchers
  • Care about the societal impact of your work and excitement about defining a new category at the intersection of AI safety and security
  • Nice to have: Experience at an AI lab, government AI safety institute, security company, or national security research organization
  • Nice to have: Experience building evaluations or benchmarks, and familiarity with frameworks such as Inspect
  • Nice to have: Familiarity with AI risk frameworks and standards, such as the OWASP Top 10 for LLM Applications, NIST AI RMF, or frontier lab safety frameworks
  • Nice to have: Security background, or experience with capture-the-flag challenges, red teaming, or cyber ranges
  • Nice to have: Experience with fine-tuning, RL environments, or building agent scaffolds

What you'll do

  • Research the behavior and failure modes of AI systems as they're actually deployed: models, RAG pipelines, agents, multi-agent systems, and the infrastructure they run on
  • Design and build custom AI evaluations end-to-end, for both AI labs and enterprises deploying AI: threat modeling, task design, environments, scoring, and reporting
  • Build the general evaluation framework that domain SMEs plug into, so a new threat domain becomes a well-formed eval suite in weeks rather than months
  • Research capability elicitation methods (scaffolding, tool use, agent harnesses, fine-tuning) so results reflect what a determined adversary could get out of a system
  • Advance the research behind our autonomous red-teaming agent: new attack techniques against models, RAG pipelines, tool calls, trust boundaries between agents, and the infrastructure AI systems touch
  • Develop methods to assess the coverage, generalization, and validity of evaluations and red-team results, including contamination, saturation, and sandbagging
  • Research mitigations for everything from prompt injection and data leakage to dangerous capabilities, misalignment, and loss of control, and turn what works into guardrails and monitoring on the platform
  • Deliver findings to customers, including the evidence they need for safety reviews and audits, and publish where it advances the field
  • Help shape product direction with the founding team, and partner with engineering and offensive security to turn research into product

Interested? Email us at careers@tinycode.io with a resume, cover note, and a link to a paper, eval, or project you're proud of.