from Hacker News

Launch HN: Shaped (YC W22) – AI-Powered Recommendations and Search

by tullie on 8/13/24, 2:19 PM with 32 comments

Hey HN! Tullie and Dan here from Shaped (https://www.shaped.ai/). We're building a semantic recommendation and search platform for marketplaces and content companies.

There’s a sandbox at https://play.shaped.ai/dashboard/home that you can use to explore demo models and evaluate results interactively. And we have a demo video at https://www.youtube.com/watch?v=toCsUYQnJ_g.

The explosion of online content, driven by both individuals and generative tools, is making it harder than ever for users to sift through the noise and find what's relevant to them. Platforms like Netflix, TikTok, and Meta have set a high bar, proving that personalized experiences are key to cutting through the clutter and engaging users effectively.

Despite advancements in AI and semantic infrastructure like vector stores, building a truly relevant recommendation or search system is still extremely difficult. It's not just about deploying the latest LLM—the difficulties lie in creating the infrastructure to orchestrate the components seamlessly. Consider the challenge of continuously fine-tuning models with fresh data while simultaneously serving real-time personalized recommendations to millions of users. It requires a delicate balancing act of speed, scale, and sophistication.

Our goal is to empower any technical team to build state-of-the-art recommendation and search systems, regardless of their data infrastructure. Here's how we eliminate the friction:

Solving Data Challenges: We integrate directly with your data sources—Segment, Amplitude, Rudderstack, and more. We handle the complexities of real-time streaming, ETLs, and data quality robustness, so you can get started in minutes.

Leveraging Cutting-Edge Models – we utilize state-of-the-art large-scale language, image, and tabular encoding models. This not only extracts maximum value from your data but also simplifies the process, even with unstructured data.

Real-time Optimization: Unlike vision or NLP tasks, recommendation system performance hinges on real-time capabilities—training, feature engineering, and serving. We've architected our platform with this at its core.

We're already helping many companies build relevant recommendations and search for their users. Outdoorsy, for example, uses us to power its RV rental marketplace. E-commerce businesses like DribbleUp and startups like Overlap have seen up to a 40% increase in both conversions and engagement when integrating Shaped.

A bit about us: Tullie was previously an AI Researcher at FAIR working on multimodal ranking at Meta. He released PyTorchVideo, a widely-used video understanding library, which contains the video understanding models that power systems like IG Reels. Dan led product research at Afterpay and Uber, driven by how behavioral psychology influences user experience.

We've been heads down building Shaped for quite a while, so this launch feels like a big milestone. We'd love to hear your feedback – technical deep dives, feature requests, you name it. Let us know what you think!

  • by candiddevmike on 8/13/24, 3:03 PM

    This seems like a tough build vs buy sell. For a lot (most?) companies, the search/recommendation system isn't necessarily optimized for the customer's search. Instead, it's a way to maximize revenue via preferred placement or inject ads. This almost always leads to a gigantic if/else chain of bespoke business analyst driven decisions for the marketplace.

    How are you going to allow folks to influence the system? Or do you see your system integrated behind their pseudo-recommendation engine?

  • by AnujNayyar on 8/13/24, 3:41 PM

    Congratulations on the launch. We've weighed up algolia, in house, type-sense etc and so I'd would have been very keen to know more, but asking for us to integrate before knowing the pricing is a tough sell.

    Would highly recommend having at least an estimated pricing calculator so we can determine if its worth our time to install.

  • by poojasengupta on 8/20/24, 7:42 AM

    Congrats (from a fellow Melbournian) on the launch. I used to work at Coles and Catch leading their online businesses, search and product recommendations was a big part of it. We had over tens of thousands of SKUs. It's harder when trying to do it with dynamic inventory locations and quantities (Coles had over 850 online stores that my team managed). We were looking at Algolia but it wasn't quite there yet (back then). I don't think anyone has solved for that as yet (I left Coles >5 years ago). Curious to hear how you would approach it.

    These days I'm the founder of a circular economy marketplace for South Asian ethnic clothing and items - PurvX. The current search is terrible (due to the low-code platform it's on), will be keeping Shaped on my radar when we re-platform.

  • by philip1209 on 8/13/24, 2:56 PM

    How do you measure quality? And, can users game that quality?

    I think that's the hardest thing on any recommendation or search system. It's really hard to do without using money as a neutral measure of value. And, without a good measure of quality - it's unclear that the system is optimizing the right metrics (without cannibalizing others).

  • by jvans on 8/13/24, 7:00 PM

    How do you personalize to the specific signals of the product, do they ingest into your infrastructure? What happens if a customer discovers a bug in a feature they're ingesting, how do they have control of retrains/pinning model versions? Who handles monitoring, the customer or your service?

  • by astronautas on 8/13/24, 4:05 PM

    How does this compare to Vespa? If the key difficulty in scaling search is infra as you say, Vespa is an interesting alternative.

  • by sidcool on 8/13/24, 3:35 PM

    What are the underlying ML models? Open source or custom trained?

  • by yding on 8/13/24, 8:49 PM

    Congrats on the launch!

  • by thierrydamiba on 8/14/24, 2:59 PM

    What vector database do you use under the hood?

  • by hajrice on 8/13/24, 3:49 PM

    How does it compare to Algolia?

  • by deepskyai on 8/13/24, 5:18 PM

    Congrats Dan and Tullie - and the rest of the team. Great to see AUSTRALIA and particularly Melbourne (formerly known as the most liveable city in the world) represented. Is there anything different now compared to what you released ~18 months ago? Or just launching on HN now?

  • by gk1 on 8/13/24, 5:10 PM

    Congrats (from Pinecone) on the launch! The e-commerce and media recommendation space desperately needs an AI-based solution without the lead-filled baggage of legacy search or recommender systems.

    > 100M+ Users I assume you mean 100M+ end-users have interacted with a site or product that uses your technology. The way it's phrased sounds like you're saying Shaped itself has 100M+ users which of course it doesn't. Consider replacing that with "100M+ interactions" or something.