Pebblebed

Pam and Keith are the firm's two most senior operators and appear to drive investment decisions alongside the wider partnership. Pam created FBLearner Flow at Facebook, was a founding member of OpenAI, and led AI at Stripe. Keith founded Facebook AI Research and served as Chief Architect at Slack. The firm's other partners — Tammie, Nicole, Michael, Kevin, and Balaji — all bring operating or venture backgrounds from Sequoia, Cruise, Facebook Reality Labs, and distributed-systems engineering.

Pebblebed's deal flow runs on the personal networks of its partners, who collectively built core AI infrastructure inside Facebook, OpenAI, Stripe, Cruise, and Slack. The firm targets 'self-protecting ideas' — startups working on technically intimidating problems that most investors lack the fluency to evaluate. That positioning draws founders who want investors that understand their compiler runtimes, robotics control loops, or distributed-training architectures without translation.

Pebblebed operates as a direct seed-stage investor, leading rounds of roughly 20 companies per fund. Its portfolio page shows only direct company investments, with no indication of fund-of-funds commitments or SPV syndications. The firm states it leads rounds, placing its capital directly on founder cap tables.

Pebblebed invests at the seed stage, occasionally entering at pre-seed for technical founders who have not yet raised institutional capital. The portfolio spans companies founded between 2017 and 2025, with investment years clustering tightly around each company's founding year. The firm does not list later-stage follow-on vehicles or growth-stage activity.

Pebblebed does not publish a formal exclusion list, but its portfolio shows no pure consumer-internet, fintech, or SaaS companies that lack a deep technical moat. Positions cluster in AI/ML tooling, robotics, computational biology, developer infrastructure, and dual-use hardware. The firm's stated thesis — backing 'self-protecting ideas' — naturally screens out copyable business models.

Pebblebed's partnership is staffed almost entirely by former senior engineers and AI infrastructure leads, not career investors. Pam and Keith built the machine-learning stacks at Facebook and OpenAI before the current AI cycle; other partners scaled teams at Cruise and Facebook Reality Labs. That operating background allows the firm to underwrite deeply technical seed-stage companies before they demonstrate standard traction metrics.

Pebblebed does not publicize a philanthropic foundation, donor-advised fund, or impact-investing carve-out. The firm's public footprint focuses exclusively on for-profit seed investing in technical founders.