Company Profile Y Combinator W25
Retrofit taught an AI to dig through vintage so you never have to scroll again
The secondhand racks are full of gems and the online experience is exhausting. Retrofit points AI agents at the pile, reads what's trending, and hands you a feed that fits your taste, size and budget.
Walk into a good vintage store in New York and the appeal is obvious - a wall of denim nobody else owns, a rack of coats with real weight, the thrill of a piece that survived thirty years and still looks right. Try to find that same thing online and the thrill evaporates. You end up ten pages deep in a resale app, filtering through blurry photos and vague descriptions, hunting for one good item in a haystack of a hundred thousand. Retrofit, a two-person company out of Y Combinator's Winter 2025 batch, is built around a simple observation: the inventory is great, the search is broken, and only one of those is worth fixing.
The company describes itself plainly - "a vintage marketplace curated by AI." What that means in practice is that Retrofit puts software between the shopper and the pile. Its AI agents read social media and sales signals to figure out what people actually want right now, then go looking for the secondhand pieces that match, pulling from partner vintage stores. The items get standardized photography and consistent, detailed descriptions. The shopper sees a clean feed tuned to their taste, size and budget instead of an ocean of listings.
The problem is the pile, not the product
Retrofit frames the pain in three parts: there is too much selection, too much of it is low quality, and finding the good stuff takes real work. Anyone who has spent a Saturday scrolling a resale app knows the feeling. The clothes exist. The patience runs out first.
There is a mirror-image problem on the other side of the counter. Independent vintage stores are sitting on inventory people would love, but listing it online is slow and tedious - photographing one-of-a-kind items, writing each description, keeping presentation consistent. Many simply can't list fast enough, and their websites get little traffic. Retrofit's pitch to those stores is that the boring part - the photographing, the tagging, the copywriting - is exactly what AI is good at.
The inventory is amazing and the experience is terrible. Retrofit is a bet that you can fix the second without touching the first.
How it actually works
Strip away the word "agent" and the pipeline is concrete. Retrofit's software behaves like a tireless trend analyst, then a personal shopper, then a merchandiser.
The Retrofit loop
A shopper can also come at it from the other direction - submit a piece of style inspiration or a specific request and let the system go find the closest thing on the racks. Free shipping and returns lower the risk of buying something you've only seen in a photo, which matters a lot when every item is unique and non-restockable.
Who's building it
Retrofit was founded in 2024 by Sandra Lifshits and Maddy Yip. Lifshits, the CEO, has a product background at ecommerce startups, a Stanford MBA and a Brown degree - but the detail that says the most about how the company thinks is that she went and volunteered at several vintage stores in New York before building. She wanted to see how the stores actually operate and how vintage shoppers actually behave. It's an unusually analog move for an AI company.
Yip brings the machine side. She worked on computer vision at Google, where she led two computer-vision patents, and studied CS at Stanford. In a company whose core job is turning a messy rack of one-of-a-kind garments into a clean, searchable feed, that's the exact expertise the problem demands.
She volunteered at vintage stores in New York before writing a product spec.On Retrofit's founding instinct
Where it sits in the market
The tailwind is real. Estimates put the US secondhand apparel market near $43 billion in 2024, with projections into the $70B-plus range by 2028 and online channels taking a growing share of that spend. The category is crowded with familiar names - Depop, Poshmark, ThredUp, The RealReal, Vestiaire Collective, eBay. What most of them share is scale without much curation: enormous inventory, thin filtering, and the shopper left to do the digging.
US secondhand apparel market (estimated)
Directional figures cited in industry coverage. Bars scaled for comparison.
FIG. 1 - A market roughly on track to triple within a decade.
Retrofit's wager is that the next wave of secondhand competition won't be won on volume. It'll be won on curation - on whether the feed has good taste. That's a harder problem in vintage than in mass retail, because you can't restock a bestseller; every item is one-of-one. Personalizing across unique inventory is precisely where Retrofit is trying to plant its flag as an early mover in applying AI agents to the space.
| Typical resale app | Retrofit | |
|---|---|---|
| Inventory shown | Everything, unfiltered | Curated to trends + taste |
| Who does the digging | You | AI agents |
| Listing quality | Seller-by-seller | Standardized photos + copy |
| Personalization | Limited | Taste, size, budget |
FIG. 2 - The pitch in one row: move the digging from the shopper to the software.
The business, plainly
Retrofit is a consumer marketplace. Shoppers buy curated pieces sourced from partner vintage stores; the stores get an automated listing channel and demand-matching they'd struggle to build themselves. It was one of a small handful of consumer companies in a YC batch otherwise thick with enterprise and developer tools, and it raised seed funding around the program in 2025. Two people, a clear problem, and a category with a decade of growth ahead of it.
The stated ambition is modest in words and large in scope: to be the first place anyone goes when they want to shop vintage online. Whether Retrofit gets there depends on something you can't fully automate - taste. The interesting bet is that you can teach software enough of it to save people the scroll.