Alex Teichman
Founder & CEO at Happenstance • Stanford CS PhD
teichman
Palo Alto, California
4.9K connections
Joined April 2025
Summary
Founder and product‑focused AI entrepreneur: Alex founded and leads Happenstance, an AI people‑search product that emphasizes natural‑language and embedding search across social and professional networks. He positions the product for founders, recruiters, and others who need deep, friend‑of‑a‑friend network search and integration into Slack/email workflows. happenstance+2
Researcher in 3D perception and robot vision: Teichman completed a Stanford CS PhD (advised by Sebastian Thrun) focused on 3D sensor‑based segmentation, tracking, and user‑trainable object recognition (STAC and group induction). His publications and dissertation show strengths in combining depth sensors with learning to make perception systems more data‑efficient and robust. stanford+2
Serial founder with hardware + software experience: Prior to Happenstance he co‑founded Lighthouse, a hardware + cloud AI camera product that applied his 3D perception research in consumer settings. Lighthouse raised institutional funding, shipped product, and — following a commercial failure in 2018 — its technology and some team members were reported to be absorbed by Apple, illustrating both technical ambition and real‑world startup experience (product, fundraising, commercialization, and shutdown). techcrunch+2
Public technical presence and community engagement: Maintains a personal research site with papers and tools, is active on X (Twitter) discussing product updates and AI engineering, and has public company / accelerator listings (YC) for Happenstance — signaling an emphasis on both technical depth and outward product/market communication. alexteichman+2
Work
Education
Projects
Writing
User‑trainable object recognition systems via group induction
January 1, 2014PhD dissertation (Stanford, 2014) describing STAC (segment, track, and classify), group induction for semi‑supervised learning from track annotations, and STRGEN algorithm for propagating segmentations — aiming to make object recognition trainable by end users with far fewer annotations.
Learning to Segment and Track in RGBD
2012-2013Conference/journal work (Teichman & Thrun and coauthors) on segmentation and tracking methods using RGB‑D data from commodity depth sensors; includes algorithmic optimizations for real‑time segmentation/tracking and applications to building labeled datasets via tracking.