Product Security Engineer

BackOps-AI•San Francisco, CA

6d•Onsite

About The Position

BackOps AI transforms supply-chain operations with agentic AI that automates complex workflows, freeing teams to focus on what matters most. Headquartered in the San Francisco Bay Area, we foster a culture of innovation, ownership, and measurable impact. Our agents take real actions inside our customers' live operations. They read from and write to the systems a supply-chain team runs its business on. That makes authorization, tenant isolation, and the blast radius of a single agent action product design questions, and it puts product security on the critical path of every enterprise deal we close. We are looking for the engineer who owns that. This is a hands-on role and the first dedicated product security hire, so you will build the product security practice and do the work at the same time. You will work closely with engineering and infrastructure on how our systems get built and run, and with product on what gets built next. You will read the code, write the fix, and ship guardrails that engineers keep switched on because they are fast and mostly right.

Requirements

5+ years in product or application security engineering, or a software engineer who moved into security and stayed hands-on.
Strong coding skills in Python, Go, TypeScript, or similar. You review real code and write the fix yourself.
Real depth in authentication, authorization, and multi-tenant isolation: OAuth and OIDC, RBAC, token and session handling, and the access-control bugs that only show up between two features.
Web and API security depth across the OWASP Top 10 classes: injection, SSRF, deserialization, and broken access control, and how these actually get exploited rather than how they are listed.
Working familiarity with the OWASP Top 10 for Agentic Applications and the OWASP Top 10 for LLM Applications. Excessive agency and prompt injection are live concerns in our product today.
Threat modeling that produces engineering work with owners and dates, not a document.
Hands-on with SAST, DAST, SCA, and secrets scanning in CI, including keeping the noise low enough that teams do not turn the checks off.
Judgment about risk. You can say what to fix now, what to accept, and why, to an audience that will push back.
Clear written and verbal communication with engineers, executives, and customers, including saying plainly what is still unknown.
Comfortable defining the structure while doing the work, in a company where requirements change week to week.

Nice To Haves

Hands-on securing LLM or agentic systems: prompt injection, insecure tool use, excessive agency and permissions, sandboxing, and handling model output as untrusted input.
SOC 2 Type I/II, ISO 27001, or similar audits taken end to end, and compliance automation tooling such as Vanta or Drata.
Enterprise security reviews and customer questionnaires answered from the engineering seat.
Multi-tenant B2B SaaS, and on-prem or customer-hosted deployment.
Running a bug bounty or coordinated disclosure program.
Security capabilities delivered as an internal product, with real adoption numbers behind them.
Standing up a product security function where none existed: charter, practices, and the culture around it.

Responsibilities

Own product security end to end: authentication, authorization, multi-tenant isolation, APIs, and the data model behind them.
Threat model the agent platform. Define what an agent is allowed to do in a customer's systems, prove the boundary holds, and design what happens when an agent is asked to do something it should not.
Build the secure development lifecycle: design review, code review standards, and scanning in CI with a triage path engineers actually use.
Run application vulnerability management end to end. Find it, prioritize it by real exploitability rather than by scanner severity, get it fixed, verify the fix.
Set the security bar for AI-assisted code. A large share of our code is written with coding agents, and human ownership of what merges is not optional.
Secure the software supply chain: dependencies, build pipeline, artifacts, and third-party integrations.
Own the engineering side of enterprise security reviews, penetration tests, and coordinated vulnerability disclosure, so customer scrutiny stops being a fire drill.
Extend the same controls to customer-hosted and on-prem deployments, where they have to hold up without our infrastructure around them.