Security Researcher

Location: San Francisco, CA or Israel · Type: Full-time, onsite

Pi Security is a product security company in San Francisco, founded by the teams who led Microsoft's vulnerability mitigation efforts and Tesla's offensive research.

We built a platform for the agentic SDLC. It changes how companies handle security vulnerabilities, not by finding more of them, but by understanding them deeply enough to fix them at the root and keep entire classes from coming back. The hard problems behind that (what to detect, how to prove a finding is real, how to remediate the way a senior engineer would) are research problems. That is where you come in.

You will lead security research at Pi. This is not a bug hunting role. Your job is to invent the approaches that become product capabilities: new ways to detect, triage, and remediate vulnerabilities using LLMs and program analysis, proven on real data before a customer ever sees them.

You own the path from idea to shipped capability. You form the hypothesis, build the proof of concept, measure whether it actually works, and partner with engineering until it is in the product. You also own the quality bar: the benchmarks and evaluations that decide whether what we ship is good enough to put in front of customers.

You will set the research agenda, not just execute one. As the team grows, you will shape how research works here.

The research agenda. Identify where new approaches can meaningfully beat the state of the art in vulnerability detection, triage, and remediation, then decide what we pursue and what we kill. No one hands you a backlog.

Approaches that ship. Build proofs of concept for new detection and remediation techniques, validate them against real world code and data, and carry the winners through to production with engineering. You are accountable for ideas becoming product, not staying research.

The quality bar. Design the datasets, benchmarks, and evaluation pipelines that measure precision, coverage, and false positive rates. Nothing reaches customers past a bar you have not signed off on, and if quality slips, you catch it first.

Vulnerability depth. Deep research into modern attack vectors across cloud (AWS/GCP), containers, microservices, APIs, AI generated code, and LLM applications. Not just how vulnerabilities are found, but how they are born, how they are fixed, and how a whole class gets eliminated.

The data foundation. The internal corpus of vulnerabilities, exploit patterns, and remediation strategies the platform learns from. Its depth and correctness are yours.

Our research voice. What we publish, where we speak, and the credibility the company earns in the security community. Your work should be visible.

Experience. 8+ years in security research, vulnerability analysis, or applied security engineering.

Startup DNA. You have worked in an early stage or 0 to 1 environment. Comfortable with ambiguity, shipping without a big org behind you, and changing direction when the data says so. This is a requirement, not a bonus.

Research to product track record. You have turned research into things that shipped: features, tools, detections in production. Not just papers or reports.

Technical depth. Deep expertise in modern application stacks (microservices, containers, cloud platforms). You understand how these systems actually break.

Builder skills. Strong programming ability in at least one modern language (Python, Go, TypeScript). Comfortable writing production quality code.

Data rigor. Experience designing experiments, building datasets or benchmarks, and measuring quality quantitatively. You do not ship on vibes.

LLM fluency. Hands on experience applying LLMs to real problems, whether evaluation, prompting, fine tuning, or agentic systems, or a demonstrated ability to get there fast.

Proven findings. A history of discovering serious vulnerabilities (CVEs welcome) and responsible disclosure.

Communication. You can explain a complex attack and its real impact clearly to engineers, executives, and customers.

Work authorization. Permanent authorization to work in the US for the San Francisco role, or in Israel for the Israel role.

We care about impact, not credentials. Things that get our attention:

• Research that went public and mattered. Publications, disclosures, or talks that changed how people think about a problem, not just filled a slot at a conference.

• A product you built at a startup. Something that shipped, that real users depended on, where you can point at your fingerprints.

• Novel ways of putting LLMs or agents to work inside real engineering or security workflows, beyond demos and prompt wrappers.

• A healthy disrespect for "that's how we've always done it," and a track record of building the better way.

You will define the research direction of a company at the stage where one person's ideas still shape the product. The team comes from top tier security organizations, the funding is in place, and the problem is real: vulnerability management is broken in ways everyone in the industry can see and almost no one is positioned to fix. If you want your research to ship and to matter, we would like to talk.