[ { "id": 245, "name": "Directed Edge", "slug": "directed-edge", "former_names": [], "small_logo_thumb_url": "/company/thumb/missing.png", "website": "http://directededge.com", "all_locations": "San Francisco, CA, USA", "long_description": "Directed Edge delivers Amazon-like \"People who bought this also bought...\" and \"We think you'd also like\"-style recommendations to other sites, using existing data being collected such as purchase or click histories. Recommendations are served in real-time via a REST API, with language bindings in several programming languages to make integration easy.", "one_liner": "Product recommendations.", "team_size": 2, "industry": "Consumer", "subindustry": "Consumer -> Home and Personal", "launched_at": 1326790627, "tags": [ "Machine Learning", "AI", "Recommendation System" ], "tags_highlighted": [], "top_company": false, "isHiring": false, "nonprofit": false, "batch": "Summer 2009", "status": "Active", "industries": [ "Consumer", "Home and Personal" ], "regions": [ "United States of America", "America / Canada" ], "stage": "Early", "app_video_public": false, "demo_day_video_public": false, "app_answers": null, "question_answers": null, "url": "https://www.ycombinator.com/companies/directed-edge", "api": "https://devasheeshg.github.io/yc-api/batches/summer-2009/directed-edge.json", "year_founded": null, "city": "San Francisco", "country": "US", "linkedin_url": null, "x_url": null, "fb_url": null, "cb_url": "https://www.crunchbase.com/organization/directed-edge", "github_url": "https://github.com/directededge", "logo_url": null, "app_video_url": null, "dday_video_url": null, "company_photos": [], "primary_partner": null, "founders": [ { "user_id": 18160, "full_name": "Scott Wheeler", "title": "Founder", "founder_bio": null, "is_active": true, "linkedin_url": null, "x_url": null, "avatar_thumb_url": "/avatars/thumb/missing.png" } ], "jobs": [], "news": [], "launches": [] }, { "id": 253, "name": "Fanchatter", "slug": "fanchatter", "former_names": [], "small_logo_thumb_url": "/company/thumb/missing.png", "website": "http://fanchatter.com", "all_locations": "Minneapolis, MN, USA", "long_description": "FanChatter fuels your fans so they can promote you better than any ad ever could.\n\nOur customers are big guys (NBC Universal, Clear Channel, Gannett) and little guys (bloggers all over the world) who pay us to source and showcase the most abundant yet least leveraged marketing material on the planet...\n\nPure, voluntary social chatter about your interest. FanChatter offers you the tools to find it, refine it, and make it multiply.\n\nAnd that is the future of advertising.", "one_liner": "FanChatter aggregates, amplifies and empowers fan content on every…", "team_size": 11, "industry": "B2B", "subindustry": "B2B -> Marketing", "launched_at": 1326790698, "tags": [ "Social", "Recommendation System" ], "tags_highlighted": [], "top_company": false, "isHiring": false, "nonprofit": false, "batch": "Summer 2009", "status": "Inactive", "industries": [ "B2B", "Marketing" ], "regions": [ "United States of America", "America / Canada" ], "stage": "Early", "app_video_public": false, "demo_day_video_public": false, "app_answers": null, "question_answers": null, "url": "https://www.ycombinator.com/companies/fanchatter", "api": "https://devasheeshg.github.io/yc-api/batches/summer-2009/fanchatter.json", "year_founded": null, "city": "Minneapolis", "country": "US", "linkedin_url": null, "x_url": null, "fb_url": null, "cb_url": "https://www.crunchbase.com/organization/fanchatter", "github_url": null, "logo_url": null, "app_video_url": null, "dday_video_url": null, "company_photos": [], "primary_partner": null, "founders": [ { "user_id": 18197, "full_name": "Marty Wetherall", "title": "Founder/CEO", "founder_bio": null, "is_active": false, "linkedin_url": "https://www.linkedin.com/in/martywetherall", "x_url": null, "avatar_thumb_url": "https://bookface-images.s3.us-west-2.amazonaws.com/avatars/e87c755478baad52466733d9aa258c9caf6e8900.jpg" }, { "user_id": 18198, "full_name": "Todd Emmel", "title": "Founder", "founder_bio": null, "is_active": false, "linkedin_url": null, "x_url": null, "avatar_thumb_url": "/avatars/thumb/missing.png" }, { "user_id": 18196, "full_name": "Luke Francl", "title": "Founder", "founder_bio": "", "is_active": false, "linkedin_url": "https://www.linkedin.com/in/lukefrancl/", "x_url": "https://twitter.com/lof", "avatar_thumb_url": "https://bookface-images.s3.us-west-2.amazonaws.com/avatars/886d0ccf8c617dccf6f313b4bfb0af4254ebaa8b.jpg" } ], "jobs": [], "news": [], "launches": [] }, { "id": 25969, "name": "Shaped", "slug": "shaped", "former_names": [ "shaped.ai Inc." ], "small_logo_thumb_url": "https://bookface-images.s3.amazonaws.com/small_logos/241b1f1931647920258ef5ae6571461ada0b62f1.png", "website": "https://shaped.ai", "all_locations": "New York City, NY, USA", "long_description": "Connect your data. Train your models. Query text, user or session context and retrieve relevant results in milliseconds. Explore our case studies to see how we’ve helped leading brands drive significant engagement and revenue: https://www.shaped.ai/case-study. ", "one_liner": "The real-time retrieval engine for search, feeds, and agents.", "team_size": 25, "industry": "B2B", "subindustry": "B2B -> Infrastructure", "launched_at": 1646119340, "tags": [ "Search", "AI", "Recommendation System" ], "tags_highlighted": [], "top_company": false, "isHiring": false, "nonprofit": false, "batch": "Winter 2022", "status": "Active", "industries": [ "B2B", "Infrastructure" ], "regions": [ "United States of America", "America / Canada" ], "stage": "Growth", "app_video_public": false, "demo_day_video_public": false, "app_answers": null, "question_answers": null, "url": "https://www.ycombinator.com/companies/shaped", "api": "https://devasheeshg.github.io/yc-api/batches/winter-2022/shaped.json", "year_founded": 2021, "city": "New York City", "country": "US", "linkedin_url": "https://www.linkedin.com/company/shapedai", "x_url": "https://twitter.com/shapedai", "fb_url": null, "cb_url": null, "github_url": null, "logo_url": "https://bookface-images.s3.amazonaws.com/small_logos/241b1f1931647920258ef5ae6571461ada0b62f1.png", "app_video_url": null, "dday_video_url": null, "company_photos": [], "primary_partner": { "name": "Brad Flora", "url": "https://www.ycombinator.com/people/brad-flora" }, "founders": [ { "user_id": 812951, "full_name": "Tullie Murrell", "title": "Founder/CEO", "founder_bio": "CEO of Shaped. \nPreviously an AI researcher at Facebook AI Research (FAIR).\nCreator of PyTorchVideo & original core contributor to PyTorch Lightning.\nPassionate about anything AI, machine-learning and statistics.", "is_active": true, "linkedin_url": "https://www.linkedin.com/in/tullie/", "x_url": "https://twitter.com/tulliemurrell", "avatar_thumb_url": "https://bookface-images.s3.us-west-2.amazonaws.com/avatars/0b422fcadf175caaedc3de64e8d33a973558dd6e.jpg" } ], "jobs": [], "news": [ { "title": "Shaped Raises Series A, Launches Self-Serve Platform to Power the Future of AI-Driven Recommendations & Search | Shaped Blog", "url": "https://www.shaped.ai/blog/shaped-raises-series-a", "date": "Jul 17, 2024" } ], "launches": [ { "id": 97225, "title": "ShapedQL: The SQL engine for search, feeds, and AI agents", "tagline": "Connect your data. Train your models. Query text, user or session context and retrieve relevant results in milliseconds.", "body": "Hi everyone 👋\\\n\\\nI'm Tullie, the founder of [Shaped](https://www.shaped.ai/).\\\n\\\n**TL;DR**\\\n\\\nToday we’re launching ShapedQL, a new language purpose built for agents, feeds and search. We built ShapedQL because we realized that while retrieval has become easier (thanks to Vector DBs), ranking and relevance are still incredibly hard.\\\n\\\nLaunch video: \\\n\\\n**Problem ❌**\\\n\\\nMost engineering teams we talk to are stuck maintaining a \"Frankenstein\" stack. To build a \"For You\" feed or give an AI Agent personalized memory, they have to glue together a vector database, a feature store (like Redis), a reranking service, and thousands of lines of Python spaghetti code.\\\n\\\n**Solution ✅**\\\n\\\nWe built ShapedQL to turn that \"house of cards\" into a single interface.\n\n![uploaded image](/media/?type=post&id=97225&key=user_uploads/812951/eed67b43-7c3b-4b0f-819a-5bbd31db2732)\n\nShapedQL is a domain-specific SQL dialect that compiles down to a high-performance, multi-stage ranking pipeline. With a single query, you can define the four stages of modern relevance:\\\n**1. Retrieve:** Fetch candidates from multiple sources (Hybrid Search, Collaborative Filtering, Trending).\\\n**2. Filter:** Apply hard constraints (e.g., \"in stock\" or \"within 50 miles\").\\\n**3. Score:** Rank results using real-time ML models (optimizing for clicks, purchases, or watch time).\\\n**4. Reorder:** Enforce diversity so your users (or Agents) don't see the same 5 items repeatedly.\n\nWe're seeing teams reduce 2,000+ lines of maintenance code down to \\~30 lines of ShapedQL, while shipping features like \"Cart Upsell\" or \"Conversational Recommendations\" in days instead of months.\n\nIf you're not a fan of SQL we also have Python or Typescript SDKs, perfect for a real-time production integration.\n\n![uploaded image](/media/?type=post&id=97225&key=user_uploads/812951/d3cc12cc-3c74-448b-adf0-ae65d9359b79)\n\n**Ask 💬**\\\n\\\n**1.** We have a Product Hunt launch at this link: . Any support or and upvote would be greatly appreciated!\\\n\\\n**2.** Please try out our interactive demo at [playground.shaped.ai](https://playground.shaped.ai), we’d love to hear any feedback on the new language. Were you able to easily make a query? Do the docs make sense? What is it missing?\\\n\\\n**3.** If the problems above resonates and you’re struggling to get true relevance from your vector store or current retrieval system, please get in touch! Book on this calendar here: and mention you’re from YC. I’ll make sure to jump on the call to help you out myself.\n\n![uploaded image](/media/?type=post&id=97225&key=user_uploads/812951/bb70b2b7-780a-4da6-a13f-511015e09776)\n\n", "url": "https://www.ycombinator.com/launches/PI9-shapedql-the-sql-engine-for-search-feeds-and-ai-agents", "votes": 15, "created_at": "2026-01-28T00:44:48.253Z" }, { "id": 88281, "title": "Shaped — Value Modeling for Search and Recommendation Systems", "tagline": "A control panel for your business objectives", "body": "# **TL;DR**\n\nWe’ve built Shaped Value Modeling, a framework that enables real-time optimization of recommendation and search ranking models for multiple objectives. It allows teams to blend different objectives (e.g. relevance, conversions, diversity) and dynamically run AB tests that drive business outcomes (with built in analytics tooling). Our community has seen awesome results, so we’re excited to bring this framework to the masses!\n\nWatch our demo [here](https://youtu.be/gYLmtMR-N4I).\n\n**Key Features:**\n\n* **Multi-objective optimization:** Combine multiple models and signals (e.g., click prediction, purchase likelihood, content quality) into a single, customizable scoring function.\n* **Dynamic ranking logic:** Define and adjust ranking objectives as your business needs change using a simple python DSL.\n* **Real-time API:** Modify ranking scores on the fly with simple API calls, enabling rapid experimentation and A/B testing.\n* **Easy to integrate:** Connect and deploy rapidly with direct integration to your existing data sources.\n* **Explainability:** Gain insights into why certain items rank higher through interpretable scoring logic, and a built in analytics platform out of the box.\n\n![uploaded image](/media/?type=post&id=88281&key=user_uploads/857360/543b423e-521a-491d-808b-0bd914f06c23)\n\n**Why This Matters:**\n\n* Rapid iteration: Experiment with new ranking strategies and adapt to changing business needs without lengthy retraining cycles.\n* Improved control: Fine-tune your ranking algorithms with precision and transparency.\n* Better alignment: Optimize for multiple KPIs simultaneously, ensuring your recommendations drive both engagement and business value.\n\n**How it Works:**\n\n1.Train separate models for different objectives (e.g., LightGBM for click prediction, BERT4Rec for purchase likelihood).\n\n2\\. Define a \"value model\" that combines these models using mathematical expressions and custom logic. Let's say we have a lightgbm click-through-rate model called ctr_model and a transformerd based next purchase model called purchase_model, you can create an ensembled value model as follows:\n\n```\nvalue_model = 0.5 * ctr_model + 0.5 * purchase_model + log(item.quality_score)\n```\n\n3\\. Dynamically adjust the value_model via API calls to optimize for different goals or user segments:\n\n```\n{\n  \"user_id\": \"83NSLX\",\n  \"config\": {\n    \"value_model\": \"0.3 * ctr_model + 0.7 * purchase_model\"\n  }\n}\n```\n\n\\\nHow our customers are using value modeling:\n\n* **Marketplaces:** Dynamically mix personalized scores with a geo-location penalizer to connect users with nearby items.\n* **E-commerce:** Balance purchase intent, engagement, and product quality.\n* **Content platforms:** Implement time-decay, boost high-quality content, and personalize recommendations.\n* **Social media:** Dynamically switch between personalized and trending models.\n\n# **Our ask**\n\nIf you know decision-makers in any of these types of organisations, we’d be super grateful for an introduction at [daniel@shaped.ai](mailto:daniel@shaped.ai)\n\n1. Marketplaces\n2. E-commerce\n3. Content platforms\n4. Social Media\n\nFor more information see our website [here](https://www.shaped.ai/value-modelling) and developer docs [here](https://docs.shaped.ai/docs/model_creation_guides/value-modeling)\n\n[Get a demo](https://meetings.hubspot.com/matt2900/meet-with-shaped-?uuid=4076ed73-5725-4ee7-ac01-2e1869237a18) if you’re interested in testing value modeling!", "url": "https://www.ycombinator.com/launches/Mxt-shaped-value-modeling-for-search-and-recommendation-systems", "votes": 8, "created_at": "2025-03-05T15:46:26.414Z" } ] }, { "id": 27734, "name": "Rubber Ducky Labs", "slug": "rubber-ducky-labs", "former_names": [ "Rubber Ducky Labs", "Inc." ], "small_logo_thumb_url": "https://bookface-images.s3.amazonaws.com/small_logos/80d1f55be06d00fd52db1ea9553e87aa6f8ac9d1.png", "website": "https://www.rubberduckylabs.io/", "all_locations": "San Francisco, CA, USA; Remote", "long_description": "We built Rubber Ducky Labs to help e-commerce teams effortlessly improve product discovery through better metadata. Our tool enables non-technical users to leverage multi-modal AI on product catalogs—just upload a CSV and start tagging metadata in minutes.", "one_liner": "AI Powered Product Discovery", "team_size": 2, "industry": "B2B", "subindustry": "B2B -> Engineering, Product and Design", "launched_at": 1671139016, "tags": [ "E-commerce", "AI", "Recommendation System" ], "tags_highlighted": [], "top_company": false, "isHiring": false, "nonprofit": false, "batch": "Winter 2023", "status": "Inactive", "industries": [ "B2B", "Engineering, Product and Design" ], "regions": [ "United States of America", "America / Canada", "Remote", "Fully Remote" ], "stage": "Early", "app_video_public": false, "demo_day_video_public": false, "app_answers": null, "question_answers": null, "url": "https://www.ycombinator.com/companies/rubber-ducky-labs", "api": "https://devasheeshg.github.io/yc-api/batches/winter-2023/rubber-ducky-labs.json", "year_founded": 2022, "city": "San Francisco", "country": "US", "linkedin_url": "https://www.linkedin.com/company/rubber-ducky-labs", "x_url": "https://twitter.com/RubberDuckyLabs", "fb_url": null, "cb_url": "https://www.crunchbase.com/organization/rubber-ducky-labs", "github_url": null, "logo_url": "https://bookface-images.s3.amazonaws.com/small_logos/80d1f55be06d00fd52db1ea9553e87aa6f8ac9d1.png", "app_video_url": null, "dday_video_url": null, "company_photos": [], "primary_partner": null, "founders": [ { "user_id": 1164034, "full_name": "Alexandra Kissel", "title": "Founder", "founder_bio": "Alexandra (née Johnson) has spent eight years in the ML tooling industry (starting with a 4-year stint as the first software engineer at @SigOpt) to help make the state of the art in machine learning effortless and easy to use. Before that, she worked on recommender systems in fashion tech. She founded Rubber Ducky Labs to be the suite of dream tools that she wished she had while working on recommender systems.", "is_active": false, "linkedin_url": "https://www.linkedin.com/in/alexandraj777/", "x_url": "https://twitter.com/alexandraj777", "avatar_thumb_url": "https://bookface-images.s3.us-west-2.amazonaws.com/avatars/026632f9ad081d2f434f06f2117ff2dd8c082677.jpg" }, { "user_id": 1298927, "full_name": "Georgia Hong", "title": "Founder", "founder_bio": "Georgia got her degree in software engineering from the University of Waterloo, during which time she completed six internships including Datadog, Cockroach Labs, Instagram and SigOpt. After graduation, she worked on Security and ML Infra at Meta.", "is_active": false, "linkedin_url": "https://www.linkedin.com/in/georgiahong/", "x_url": "", "avatar_thumb_url": "https://bookface-images.s3.us-west-2.amazonaws.com/avatars/ed1edaba5f6e87a1c69d598bcd6d1dc7233ac6a9.jpg" } ], "jobs": [], "news": [ { "title": "The quest for better AI recommendation engines | TechCrunch", "url": "https://techcrunch.com/2023/06/27/rubber-ducky-labs-keeps-ai-engines-afloat/", "date": "Jun 27, 2023" } ], "launches": [ { "id": 69112, "title": "Rubber Ducky Labs: Stop recommending ski jackets in June", "tagline": "Rubber Ducky Labs builds better recommender systems with machine learning plus human expertise", "body": "![uploaded image](/media/?type=post&id=69112&key=user_uploads/1164034/d71009b4-6b77-430f-a071-18fbccf522c5)\n\n**TL / DR;** Rubber Ducky Labs helps e-commerce companies avoid tone-deaf product recommendations — like featuring ski jackets in June. It works by combining machine learning with human expertise to produce the best product recommendations.\n\n![uploaded image](/media/?type=post&id=69112&key=user_uploads/1164034/3234f8dd-48e2-4edc-8525-f63699ee47e9)\n\n# Introduction\n\nCo-founders Alexandra and Georgia have over a decade of experience shipping products in fashion tech, B2B ML tooling, and high-scale data infrastructure, and are passionate about making the state of the art in machine learning effortless and easy to understand.\n\nAlexandra’s career started with her dream job working on clothing recommender systems for fashion tech companies, and she and Georgia founded Rubber Ducky Labs to be the suite of dream tools that they wished were available for building recommender systems.\n\n# The Problem\n\nYou’re a PM at a fashion tech startup. Your app is recommending that you buy ski jackets. It’s June.\n\nLooks like nobody told the algorithm that summer isn’t ski jacket season, and now it’s your job to figure out whether or not your algorithm will start driving away users. If your team’s metrics dashboards can’t answer the question, you’ll need to file a ticket for the data science team to run some analysis. Looks like it’s going to be at least a week before you have an answer.\n\n# The Solution\n\nWith Rubber Ducky Labs you’ll have an answer within minutes. Now, you’ll be able to understand whether you looked at a lot of ski jackets recently, so it’s just happening to you, or if your colleague recently flipped a switch to promote a brand that happens to carry a lot of ski jackets.\n\n![uploaded image](/media/?type=post&id=69112&key=user_uploads/1164034/918010a5-c9d2-49f6-b1a0-44086fb79f57)\n\nOur solution was developed from over a hundred conversations to answer the question of “what’s going on with my recommender system?” With Rubber Ducky Labs you can:\n\n* Explore your data visually - with product images\n* Drill down from aggregate metrics all of the way to individual item or user journeys\n* Consolidate and debug your company’s business logic\n* View annual and seasonal trends\n* Facet, segment, and filter and your data\n* View side-by-side comparisons of models or ranking changes\n\n# The Vision\n\nThe mission of Rubber Ducky Labs is to allow people with domain expertise — like product managers, merchandisers, marketers, growth hackers, and founders — to easily build intuition about and incorporate domain knowledge into their recommender systems.\n\nToday’s launch is just the first step. We want to help you do everything from consolidating business logic to previewing side by side model comparison to launching production experiments, all within RDL’s tools.\n\nIf this sounds like something you’re interested in, we’d love to build together.\n\n# Technical Details\n\nSetup took 90 minutes of time with our first user. Under the hood, Rubber Ducky Labs is a fully hosted web app that connects directly to your data warehouse on the backend to pull your custom metrics and data. No need to deploy any services or change any code, plus you get to bring your own model! Additionally, the web and API are authenticated with Auth0 to keep your data safe.\n\n# Our Ask\n\nWe’d love to learn more about your recommender system, and start identifying areas for improvement. Interested? Email [alexandra@rubberduckylabs.io](mailto:alexandra@rubberduckylabs.io) to get started.\n\nHave a friend who works with recommender systems? We’d love to meet them! Email us at [founders@rubberduckylabs.io](mailto:founders@rubberduckylabs.io).", "url": "https://www.ycombinator.com/launches/Hyi-rubber-ducky-labs-stop-recommending-ski-jackets-in-june", "votes": 52, "created_at": "2023-02-08T00:47:04.990Z" } ] } ]