Petrarch operates a marketplace for proprietary datasets that sit inside enterprises and legacy industries, buying and licensing that data to AI companies that need training material beyond what the open web offers. According to the company's own description, it sources enterprise-facing datasets from distressed businesses and legacy industries, clears and anonymizes them, then brokers them to AI companies building for the physical world. Early transactions have included construction plans, autonomous driving recordings, and codebases, which points at buyers working on robotics, self-driving, and code generation models.
The launch matters because the training data supply story has shifted. Frontier labs have largely exhausted the open internet, and the next wave of gains depends on domain-specific corpora that live in private hands, from CAD files to sensor logs to internal software repositories. Petrarch is positioning itself as the intermediary that finds these assets, handles the legal and privacy work to make them sellable, and matches them with labs willing to pay.
The team is Ian Lee, Samuel Hahn, and Sudhish Swain, three Harvard dropouts with prior experience at Scale and Meridian AI. Petrarch is part of Y Combinator's Summer 2026 batch, works with YC partner Vivian Midha Shen, and is based in San Francisco with a team of three.
Comments (7)
Sign in to join the discussion.
D
Dara OkonkwoJul 13, 2026
A market maker implies liquidity on both sides. Who's the counterparty when the data is locked in a Fortune 500 legal department that hasn't approved a vendor since 2019?
L
Linh TranJul 13, 2026
Curious how you're pricing datasets that have zero comps. Is it auction-based, negotiated, or vibes-based tier 1 through tier 3?
P
Priya VatsalJul 13, 2026
Three Harvard dropouts is basically the required starter kit now. At this point YC should just issue the diploma-shredder in the welcome package.
M
matteo b.Jul 13, 2026
The wordmark has that serif-with-a-secret vibe and I respect it, but the announcement graphic is doing a lot of gradient work for a company named after a 14th century poet.
K
Kenji O.Jul 13, 2026
The launch video moves fast but the voiceover buries the actual pitch until about 22 seconds in. If you're selling to enterprise buyers, put the words market maker on screen at second one.
Z
zaraJul 13, 2026
This might genuinely be one of the last data infra companies to launch without an agent doing the pitch itself. Feels like watching the last flip phone ship.
V
Vik ShevchenkoJul 13, 2026
Reminds me a bit of what Scale did early on, but on the supply side instead of labeling. If they crack procurement, the moat writes itself.
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
Petrarch
The first market maker for specialized data
The first market maker for specialized data
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
Petrarch operates a marketplace for proprietary datasets that sit inside enterprises and legacy industries, buying and licensing that data to AI companies that need training material beyond what the open web offers. According to the company's own description, it sources enterprise-facing datasets from distressed businesses and legacy industries, clears and anonymizes them, then brokers them to AI companies building for the physical world. Early transactions have included construction plans, autonomous driving recordings, and codebases, which points at buyers working on robotics, self-driving, and code generation models. The launch matters because the training data supply story has shifted. Frontier labs have largely exhausted the open internet, and the next wave of gains depends on domain-specific corpora that live in private hands, from CAD files to sensor logs to internal software repositories. Petrarch is positioning itself as the intermediary that finds these assets, handles the legal and privacy work to make them sellable, and matches them with labs willing to pay. The team is Ian Lee, Samuel Hahn, and Sudhish Swain, three Harvard dropouts with prior experience at Scale and Meridian AI. Petrarch is part of Y Combinator's Summer 2026 batch, works with YC partner Vivian Midha Shen, and is based in San Francisco with a team of three.
Comments (7)
A market maker implies liquidity on both sides. Who's the counterparty when the data is locked in a Fortune 500 legal department that hasn't approved a vendor since 2019?
Curious how you're pricing datasets that have zero comps. Is it auction-based, negotiated, or vibes-based tier 1 through tier 3?
Three Harvard dropouts is basically the required starter kit now. At this point YC should just issue the diploma-shredder in the welcome package.
The wordmark has that serif-with-a-secret vibe and I respect it, but the announcement graphic is doing a lot of gradient work for a company named after a 14th century poet.
The launch video moves fast but the voiceover buries the actual pitch until about 22 seconds in. If you're selling to enterprise buyers, put the words market maker on screen at second one.
This might genuinely be one of the last data infra companies to launch without an agent doing the pitch itself. Feels like watching the last flip phone ship.
Reminds me a bit of what Scale did early on, but on the supply side instead of labeling. If they crack procurement, the moat writes itself.