In this Data Driven NYC presentation, Cheryl Liu, founder and CEO of seed‑stage startup Raspberry AI, outlines the company’s mission to help fashion brands create designs that consumers love by leveraging AI. She explains the core industry challenges: 40% of fashion inventory goes unsold, driven by rapidly shifting Gen Z and millennial tastes sourced from social media, and long 12‑18‑month design‑to‑manufacture cycles that cause brands to miss trends. Liu cites that 73% of fashion executives plan to adopt AI this year. Raspberry AI’s product suite addresses the full design workflow: data‑driven trend recommendations from Google search and TikTok engagement; Sketch‑to‑Render that turns sketches into photorealistic images in seconds, saving up to 30% on sample costs; an Image Mixer that preserves silhouette while blending fabrics and applying novel textures; instant color and print editing via text prompts; brush‑based direct image edits; AI‑generated on‑body photography for virtual try‑ons; and conversion of product photos into 2‑D technical drawings for manufacturing. She demonstrates each feature with short video demos, emphasizing speed‑to‑market, cost reduction, waste minimisation, and higher conversion e‑commerce assets. The talk concludes with a brief hiring call for machine‑learning and product talent.
Chapters
Hello everyone, I'm Cheryl Liu. I'm the founder and CEO of Raspberry AI. We're a seed stage startup based here in New York and in the Bay Area. Thank you. So our mission is really simple. It is to help brands create designs that people love. And while it sounds really simple, in practice it's actually very, very difficult, evidenced by the fact that 40% of everything created in fashion goes unsold. So next time you go to a store, just imagine 40% of the things in the store are never sold, never used. And why is that? It's because the consumer purchase behaviors have changed so radically in the last 10 years. But brands haven't been able to keep up with that change. One root cause is that almost 80% of Gen Z and millennials, for us breaking everything, we get style inspiration on social media and online searching on Google, and brands aren't able to leverage and process that huge amount of really rich consumer data that's out there. So they're just kind of blindly guessing what to design. They're using their gut feeling, but there's so much data that they could be taking advantage of. The second reason all that inventory goes unsold is because it hits the market oftentimes too late. So typically it takes brands 12 to 18 months to design and manufacture their products. You can imagine if you're designing something today to be sold in 18 months, consumer trends could have completely changed and people want different things. So AI is going to help brands make huge waves on this. Actually, fashion is one of the biggest industries that is expected to have one of the biggest positive impacts to the profits and revenues, but also the waste. So fashion executives are super excited to adopt the technology, and actually 73% of them are adopting it this year. You guys have probably seen AI-generated artwork designs out there. Very cool, very beautiful, but One thing we're really excited about at Raspberry is how do we turn those cool designs into real business results, improving revenue, improving speed to market, cutting down sample cost, and ultimately reducing waste and returns. So everything you see here on the slide is something we help the brands with for their end-to-end design process, from the initial customer market research to the ideation. This is the more traditional, you know, prompt your design or put in a reference image. Also, we have prototyping and editing features in our suite, as well as the ability to convert product images into 2D technical drawings. Also in beta with some of our customers is the ability to do AI focus groups and customer testing and producing those high-conversion product descriptions and styled photography that you see on their e-commerce sites. So the next couple slides are a few really quick video demos. This is the first part of our product suite. We take all this really rich Google search data, what people are really searching for on Google, what they're engaging with on TikTok, what the assortment's like online, and then make really targeted data-driven recommendations on what people actually want to buy. We can also instantly change designs with a few words. So upload a reference image. Let's say we want to change this rhinestone boot into a leopard print boot. We just Simply type that in and then you get beautiful results that honor the original silhouette. This is the Sketch to Render feature. This is what saves brands up to 30% on sample cost. So what you're seeing here is you upload a sketch of an image that you wanna produce and then it turns it into a photorealistic render in 5 seconds just by typing in the words black leather booty. So they don't have to then physically create those samples. One of our other most popular features is an image mixer that allows the brands to have full silhouette control. So I, I love Midjourney, love DALL·E, ChatGPT, use it all the time. This is just to illustrate how we've optimized it specifically for the fashion use case. So here we're uploading a reference image of a puff sleeve dress and tree bark that we're inspired by. So you can see Raspberries is the one that really honors that original silhouette you see there. Using the prompt dress made out of tree bark. And our image mixer also accurately blends fabrics together. So let's say you're really, you really like this tinsel fabric and you wanted to see it applied on these very cool pants here with the slits. And yeah, I hope someday these go on the market. Another product feature we have, which is very popular, is the ability to instantly add colors and prints. And apply exactly what the end user is looking for. Control is obviously a huge part of generative AI challenges we face. And adding on to the control piece, we actually allow our clients to make direct edits onto images. So for example, they can brush the exact area that they want to change. So for example, if they wanted to add long sleeves to this dress, could just brush over that and type in, add long sleeves to the dress. So it really helps them visualize all sorts of combinations in like 10 seconds. With the on-body photography, you know, photo shoots can cost $10K to $50K each, and with AI you can create this beautiful pre-visualization in any background. You can have them on the moon, you can have any size, any hairstyle. And then finally, we also allow brands to transform a product photo into a technical drawing, which they can then further refine. And this is a really important asset for their manufacturing process. And here's the just quick comparison for the, uh, technical drawing. We've trained all of our models, um, in-house and hopefully are giving the results that our clients like. And if anyone is looking to join a seed stage startup, we are hiring for machine learning engineers, product and this stuff. Thank you so much.
