About Solumn AI
Solumn AI builds the safety and alignment layer for frontier AI systems. We work with leading AI labs to make advanced models deployment-ready across red teaming, vulnerability discovery, safety evaluations, agent-behavior testing, and the datasets and frameworks that make it measurable. If capability asks "what can the model do?", we ask "what happens when it fails, and how do we prevent it?"
Every technology cycle begins with capability and matures through trust. AI is crossing that threshold now, and the assurance layer is becoming foundational. We're building the firm that does it to a benchmark standard.
Role Overview
We're hiring a Technical Program Manager to own delivery across both sides of our work: the client engagements we run for labs, and the proactive eval datasets we build, publish, and use to drive outreach and credibility.
What You'll Own
Client delivery
- Own end-to-end delivery on active lab engagements — scoping, timelines, milestones, quality bar, and handoffs
- Translate lab safety team requirements into clear briefs for our researchers and engineers, and communicate our work back to technical counterparts at frontier labs
- Be the primary point of contact for clients through an engagement — someone who can hold a substantive conversation about what we're building and why
Proactive eval and research work
- Drive our pipeline of self-initiated eval datasets — work we build ahead of any client request, publish externally, and use to open conversations with labs
- Coordinate across research and engineering to keep this work moving alongside client commitments — scoping, tracking progress, managing publication and release
- Help shape which proactive bets we take: which failure modes are timely, which categories have the best opening for us, and what the outreach story should be
Operations
- Build delivery infrastructure as we scale — how we scope, how we QA, how we improve across engagements
- Track the full pipeline across client and proactive work, and surface risks early
What We're Looking For
- 4–7 years in technical program management at a research org, AI lab, or technical services team
- Working knowledge of how frontier models are trained — the kinds of datasets, evaluation environments, and RL setups labs actually use, and what makes each useful or not. You don't need to have built these yourself, but you should have been close enough to the work to have opinions about it
- Strong technical fluency to engage with researchers and engineers on substance — understanding what makes an eval rigorous, when a scope is too broad, and when a dataset is actually solving the problem it's meant to solve
- Comfortable in client-facing conversations with technical stakeholders — safety researchers and engineers at frontier labs
- Ownership orientation that goes beyond coordination — you notice what needs to happen and make it happen
Nice to Have
- Familiarity with the AI safety and evaluation landscape — the key organizations, frameworks, and research directions
- Experience at a startup — particularly if you've had to manage complex technical work with limited structure around you
