Wyvern is an end-to-end platform to help marketplaces build and scale custom machine learning pipelines, with the goal of enabling data scientists to own the full machine learning stack. It offers a bundle of:\n1. Feature store to store and serve features\n2. Model service to serve ML models \n3. Feature + Model observability platform \n4. Orchestration framework for ML Pipelines
While being mindful of the following constraints:\n1. Pipeline evaluation in < 200ms to optimize user experience [1]\n2. Minimizing train / test skew for model training
Suchintan built the ML Platform at Faire and Gopuff to improve their Search and Discovery experience. At both places, the platform became an engine that empowered the data team to independently deliver new models to production, generating over $100M of impact.
Small marketplaces usually buy solutions like Algolia and AWS Personalize for Search and Discovery. Once a marketplace grows to a certain size (usually >$100M GMV), they hit limitations around how many custom (\u201dlong-tail\u201d) insights they can incorporate into these solutions.
Long-tail insights may come in the form of \u201cfeatures\u201d that help move model accuracy, or changes in model objectives that help optimize for underlying business objectives:\n- B2B Marketplaces like Faire are able to segment users based on what they\u2019re selling through on their storefront by asking them to categorize themselves on sign-up, and feed that into machine learning models to cold-start personalization for new users.\n- Margin-sensitive companies like Gopuff are able to build a function to predict the expected revenue of a product being shown to the user, and ordered the results accordingly. This was composed by several ML models coming together trying to predict the probability the user would purchase a product if we showed it to them, and the expected margin of that transaction. Cart-state as a signal is also very helpful when ranking complementary products for users, ie specific types of chips would be favored if you had some coke in your cart, versus sprite
We\u2019ve talked to a bunch of other marketplaces and we\u2019ve learned they have their own long-tail insights that may improve the user experience within their own models:\n- Recipe marketplaces like Cookpad may associate each recipe with a flavour profile, and leverage that data to map recipe flavors to predicted user profiles (ie weight Savoury recipes higher for users that enjoy Savoury food)\n- Device reselling marketplaces like Valyuu may use the brand of the users\u2019 device they\u2019re viewing the website on to predict which type of product they were most likely to purchase (iPhone users buying another iPhone)
The question we asked ourselves is: How much engineering involvement is actually required here, and how much can we generalize? We built Wyvern to abstract ML engineering work mentioned above away from data scientists. As a result, they can just focus on defining the request/response of the API, the model, the features for the model, the business logic, and finally training the models with the feedback data generated by the ML pipeline.
We would love your opinions and hot-takes. Please follow our Quickstart (https://github.com/Wyvern-AI/wyvern#quickstart) to give it a try.
[1] https://iarapakis.github.io/papers/TOIS17.pdf"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Wyvern \u2013 Real-time machine learning platform for marketplaces"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/Wyvern-AI/wyvern"}},"_tags":["story","author_wintonzheng","story_37501070","show_hn"],"author":"wintonzheng","created_at":"2023-09-13T20:06:47Z","created_at_i":1694635607,"num_comments":0,"objectID":"37501070","points":15,"story_id":37501070,"story_text":"Hello HN. I\u2019m Shu, here with my best friend Suchintan. We\u2019re the creators of Wyvern (https://github.com/Wyvern-AI/wyvern), a real-time machine learning framework to help marketplaces adopt ML earlier in their life.
Wyvern is an end-to-end platform to help marketplaces build and scale custom machine learning pipelines, with the goal of enabling data scientists to own the full machine learning stack. It offers a bundle of:\n1. Feature store to store and serve features\n2. Model service to serve ML models \n3. Feature + Model observability platform \n4. Orchestration framework for ML Pipelines
While being mindful of the following constraints:\n1. Pipeline evaluation in < 200ms to optimize user experience [1]\n2. Minimizing train / test skew for model training
Suchintan built the ML Platform at Faire and Gopuff to improve their Search and Discovery experience. At both places, the platform became an engine that empowered the data team to independently deliver new models to production, generating over $100M of impact.
Small marketplaces usually buy solutions like Algolia and AWS Personalize for Search and Discovery. Once a marketplace grows to a certain size (usually >$100M GMV), they hit limitations around how many custom (\u201dlong-tail\u201d) insights they can incorporate into these solutions.
Long-tail insights may come in the form of \u201cfeatures\u201d that help move model accuracy, or changes in model objectives that help optimize for underlying business objectives:\n- B2B Marketplaces like Faire are able to segment users based on what they\u2019re selling through on their storefront by asking them to categorize themselves on sign-up, and feed that into machine learning models to cold-start personalization for new users.\n- Margin-sensitive companies like Gopuff are able to build a function to predict the expected revenue of a product being shown to the user, and ordered the results accordingly. This was composed by several ML models coming together trying to predict the probability the user would purchase a product if we showed it to them, and the expected margin of that transaction. Cart-state as a signal is also very helpful when ranking complementary products for users, ie specific types of chips would be favored if you had some coke in your cart, versus sprite
We\u2019ve talked to a bunch of other marketplaces and we\u2019ve learned they have their own long-tail insights that may improve the user experience within their own models:\n- Recipe marketplaces like Cookpad may associate each recipe with a flavour profile, and leverage that data to map recipe flavors to predicted user profiles (ie weight Savoury recipes higher for users that enjoy Savoury food)\n- Device reselling marketplaces like Valyuu may use the brand of the users\u2019 device they\u2019re viewing the website on to predict which type of product they were most likely to purchase (iPhone users buying another iPhone)
The question we asked ourselves is: How much engineering involvement is actually required here, and how much can we generalize? We built Wyvern to abstract ML engineering work mentioned above away from data scientists. As a result, they can just focus on defining the request/response of the API, the model, the features for the model, the business logic, and finally training the models with the feedback data generated by the ML pipeline.
We would love your opinions and hot-takes. Please follow our Quickstart (https://github.com/Wyvern-AI/wyvern#quickstart) to give it a try.
[1] https://iarapakis.github.io/papers/TOIS17.pdf","title":"Show HN: Wyvern \u2013 Real-time machine learning platform for marketplaces","updated_at":"2024-09-20T15:06:20Z","url":"https://github.com/Wyvern-AI/wyvern"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"borja_d"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"Hi HN,
As PMs (Google, Gopuff, Shiji), my cofounder and I ran hundreds of user tests over the years. And every time it felt broken:\nIt could take 2\u20135 days just to get enough usable results
We spent hours watching recordings to extract a handful of insights
\u201cProfessional testers\u201d rushed through tasks for pay, creating biased feedback
Platforms often started at \u20ac10k+ per year, with hidden fees on top
The result: slow iteration cycles, unreliable feedback, and user testing that often felt like a tax rather than a tool.\nWe started asking: what if AI could help?\nCould synthetic users replicate realistic human behaviors?
Could we simulate thousands of testers instantly instead of recruiting them?
Would that make user testing accessible to any team, not just those with big budgets?
That exploration led us to build Uxia, an AI-powered user testing tool that:\nDelivers actionable insights in ~5 minutes, not days
Uses AI profiles to simulate thousands of behaviors
Offers flat pricing \u2192 unlimited tests, unlimited users, no hidden costs
You can upload a prototype, design, or flow and see where synthetic testers get stuck, what paths they take, and how they interact \u2014 all without waiting on recruiting or biased feedback loops.\nOf course, we know this approach isn\u2019t perfect. Synthetic users won\u2019t fully replace human intuition, but we think they can remove friction from the early stages of iteration and help teams test much more often. We\u2019re also on PH today if you want to support the launch).\nWe\u2019d love your feedback:
Where do you think AI-driven testers could work well, and where would they fall short?
Would you trust synthetic feedback enough to guide real product decisions?
If you\u2019ve struggled with user testing, what\u2019s the one thing you wish could be different?
Thanks for reading, and happy to answer anything, we\u2019ll be around all day."},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Uxia: AI-powered user testing in minutes"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.uxia.app/"}},"_tags":["story","author_borja_d","story_45156152","show_hn"],"author":"borja_d","children":[45156162,45156167,45159980],"created_at":"2025-09-07T07:31:10Z","created_at_i":1757230270,"num_comments":4,"objectID":"45156152","points":3,"story_id":45156152,"story_text":"Hi HN,
As PMs (Google, Gopuff, Shiji), my cofounder and I ran hundreds of user tests over the years. And every time it felt broken:\nIt could take 2\u20135 days just to get enough usable results
We spent hours watching recordings to extract a handful of insights
\u201cProfessional testers\u201d rushed through tasks for pay, creating biased feedback
Platforms often started at \u20ac10k+ per year, with hidden fees on top
The result: slow iteration cycles, unreliable feedback, and user testing that often felt like a tax rather than a tool.\nWe started asking: what if AI could help?\nCould synthetic users replicate realistic human behaviors?
Could we simulate thousands of testers instantly instead of recruiting them?
Would that make user testing accessible to any team, not just those with big budgets?
That exploration led us to build Uxia, an AI-powered user testing tool that:\nDelivers actionable insights in ~5 minutes, not days
Uses AI profiles to simulate thousands of behaviors
Offers flat pricing \u2192 unlimited tests, unlimited users, no hidden costs
You can upload a prototype, design, or flow and see where synthetic testers get stuck, what paths they take, and how they interact \u2014 all without waiting on recruiting or biased feedback loops.\nOf course, we know this approach isn\u2019t perfect. Synthetic users won\u2019t fully replace human intuition, but we think they can remove friction from the early stages of iteration and help teams test much more often. We\u2019re also on PH today if you want to support the launch).\nWe\u2019d love your feedback:
Where do you think AI-driven testers could work well, and where would they fall short?
Would you trust synthetic feedback enough to guide real product decisions?
If you\u2019ve struggled with user testing, what\u2019s the one thing you wish could be different?
Thanks for reading, and happy to answer anything, we\u2019ll be around all day.","title":"Show HN: Uxia: AI-powered user testing in minutes","updated_at":"2026-03-05T22:40:24Z","url":"https://www.uxia.app/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"mfiguiere"},"title":{"matchLevel":"none","matchedWords":[],"value":"The Fantasy of Instant Delivery is Imploding"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"https://www.bloomberg.com/news/features/2022-10-24/gopuff-layoffs-signal-instant-delivery-s-demise"}},"_tags":["story","author_mfiguiere","story_33326666"],"author":"mfiguiere","children":[33326746,33328526,33329926,33332487],"created_at":"2022-10-25T05:41:58Z","created_at_i":1666676518,"num_comments":11,"objectID":"33326666","points":15,"story_id":33326666,"title":"The Fantasy of Instant Delivery is Imploding","updated_at":"2024-09-20T12:22:04Z","url":"https://www.bloomberg.com/news/features/2022-10-24/gopuff-layoffs-signal-instant-delivery-s-demise"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"jocker12"},"title":{"matchLevel":"none","matchedWords":[],"value":"The Fantasy of Instant Delivery Is Imploding"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"https://www.bloomberg.com/news/features/2022-10-24/gopuff-layoffs-signal-instant-delivery-s-demise"}},"_tags":["story","author_jocker12","story_33408441"],"author":"jocker12","children":[33409024,33409773,33410132,33410145,33413307],"created_at":"2022-10-31T16:43:34Z","created_at_i":1667234614,"num_comments":5,"objectID":"33408441","points":9,"story_id":33408441,"title":"The Fantasy of Instant Delivery Is Imploding","updated_at":"2024-09-20T12:21:04Z","url":"https://www.bloomberg.com/news/features/2022-10-24/gopuff-layoffs-signal-instant-delivery-s-demise"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"walterbell"},"title":{"matchLevel":"none","matchedWords":[],"value":"Can the Mad Rush to Deliver Your Groceries in 10 Minutes Be Profitable?"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"https://www.bloomberg.com/news/articles/2021-08-02/gopuff-gorillas-getir-aim-to-deliver-groceries-in-10-minutes"}},"_tags":["story","author_walterbell","story_28037088"],"author":"walterbell","children":[28038757],"created_at":"2021-08-02T14:38:38Z","created_at_i":1627915118,"num_comments":1,"objectID":"28037088","points":3,"story_id":28037088,"title":"Can the Mad Rush to Deliver Your Groceries in 10 Minutes Be Profitable?","updated_at":"2024-09-20T09:04:39Z","url":"https://www.bloomberg.com/news/articles/2021-08-02/gopuff-gorillas-getir-aim-to-deliver-groceries-in-10-minutes"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"taubek"},"title":{"matchLevel":"none","matchedWords":[],"value":"Companies That Raised Money in 2021 Are More Likely to See Layoffs"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"https://news.crunchbase.com/layoffs/companies-raised-funding-2021-gopuff-chime/"}},"_tags":["story","author_taubek","story_34518585"],"author":"taubek","created_at":"2023-01-25T14:26:37Z","created_at_i":1674656797,"num_comments":0,"objectID":"34518585","points":2,"story_id":34518585,"title":"Companies That Raised Money in 2021 Are More Likely to See Layoffs","updated_at":"2024-09-20T13:09:42Z","url":"https://news.crunchbase.com/layoffs/companies-raised-funding-2021-gopuff-chime/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"jkuria"},"title":{"matchLevel":"none","matchedWords":[],"value":"What Will Work\u2013and What Won\u2019t\u2013In the Sprint for Fast Delivery"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"https://www.wsj.com/articles/gopuffs-co-ceo-on-what-will-workand-what-wontin-the-sprint-for-quick-delivery-11651762617"}},"_tags":["story","author_jkuria","story_31443066"],"author":"jkuria","created_at":"2022-05-20T03:35:24Z","created_at_i":1653017724,"num_comments":0,"objectID":"31443066","points":1,"story_id":31443066,"title":"What Will Work\u2013and What Won\u2019t\u2013In the Sprint for Fast Delivery","updated_at":"2024-09-20T11:09:42Z","url":"https://www.wsj.com/articles/gopuffs-co-ceo-on-what-will-workand-what-wontin-the-sprint-for-quick-delivery-11651762617"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"pseudolus"},"title":{"matchLevel":"none","matchedWords":[],"value":"Groceries in 10 minutes: NYC has become a testing ground for ultrafast delivery"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"https://www.cnbc.com/2021/10/21/gopuff-gorillas-and-others-flood-new-york-with-instant-delivery-options.html"}},"_tags":["story","author_pseudolus","story_28942812"],"author":"pseudolus","created_at":"2021-10-21T11:10:14Z","created_at_i":1634814614,"num_comments":0,"objectID":"28942812","points":1,"story_id":28942812,"title":"Groceries in 10 minutes: NYC has become a testing ground for ultrafast delivery","updated_at":"2024-09-20T09:39:46Z","url":"https://www.cnbc.com/2021/10/21/gopuff-gorillas-and-others-flood-new-york-with-instant-delivery-options.html"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"zalmoxes"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"GoBuffalo a New Go Web Framework"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"http://gobuffalo.io/docs/getting-started"}},"_tags":["story","author_zalmoxes","story_13332940"],"author":"zalmoxes","created_at":"2017-01-06T00:50:33Z","created_at_i":1483663833,"num_comments":0,"objectID":"13332940","points":2,"story_id":13332940,"title":"GoBuffalo a New Go Web Framework","updated_at":"2024-09-20T00:11:57Z","url":"http://gobuffalo.io/docs/getting-started"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"B-Scan"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"How to Deploy GoBuffalo on Google's App Engine"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"https://golangtesting.com/posts/gobuffalo-app-engine"}},"_tags":["story","author_B-Scan","story_19806753"],"author":"B-Scan","created_at":"2019-05-02T11:31:58Z","created_at_i":1556796718,"num_comments":0,"objectID":"19806753","points":1,"story_id":19806753,"title":"How to Deploy GoBuffalo on Google's App Engine","updated_at":"2024-09-20T04:09:02Z","url":"https://golangtesting.com/posts/gobuffalo-app-engine"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"sjellis"},"title":{"matchLevel":"none","matchedWords":[],"value":"Buffalo \u2013 MVC Web Framework for Go"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"http://gobuffalo.io"}},"_tags":["story","author_sjellis","story_13387059"],"author":"sjellis","children":[13387268,13387540,13387564,13388002,13388334,13388788,13389013,13389160,13390175,13390952,13391185,13454022],"created_at":"2017-01-12T23:20:11Z","created_at_i":1484263211,"num_comments":43,"objectID":"13387059","points":167,"story_id":13387059,"title":"Buffalo \u2013 MVC Web Framework for Go","updated_at":"2025-01-04T11:00:02Z","url":"http://gobuffalo.io"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"scns"},"title":{"matchLevel":"none","matchedWords":[],"value":"A Go web development eco-system, designed to make your life easier"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"https://gobuffalo.io"}},"_tags":["story","author_scns","story_36616679"],"author":"scns","created_at":"2023-07-06T14:42:14Z","created_at_i":1688654534,"num_comments":0,"objectID":"36616679","points":4,"story_id":36616679,"title":"A Go web development eco-system, designed to make your life easier","updated_at":"2024-09-20T14:32:17Z","url":"https://gobuffalo.io"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"sjellis"},"title":{"matchLevel":"none","matchedWords":[],"value":"Buffalo v0.8.0 \u2013 MVC Web for Go"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"https://blog.gobuffalo.io/buffalo-v0-8-0-released-33f1786eadf"}},"_tags":["story","author_sjellis","story_14005464"],"author":"sjellis","children":[14005479],"created_at":"2017-03-31T16:34:25Z","created_at_i":1490978065,"num_comments":1,"objectID":"14005464","points":2,"story_id":14005464,"title":"Buffalo v0.8.0 \u2013 MVC Web for Go","updated_at":"2024-09-20T00:32:32Z","url":"https://blog.gobuffalo.io/buffalo-v0-8-0-released-33f1786eadf"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"elliotlarson"},"title":{"matchLevel":"none","matchedWords":[],"value":"Buffalo \u2013 a new Go web framework"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"http://gobuffalo.io/docs/getting-started"}},"_tags":["story","author_elliotlarson","story_13385150"],"author":"elliotlarson","children":[13385273],"created_at":"2017-01-12T19:11:25Z","created_at_i":1484248285,"num_comments":1,"objectID":"13385150","points":2,"story_id":13385150,"title":"Buffalo \u2013 a new Go web framework","updated_at":"2024-09-20T00:17:28Z","url":"http://gobuffalo.io/docs/getting-started"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"todd3834"},"title":{"matchLevel":"none","matchedWords":[],"value":"A Go web development eco-system, designed to life easier"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"https://gobuffalo.io/"}},"_tags":["story","author_todd3834","story_20485830"],"author":"todd3834","created_at":"2019-07-20T14:07:32Z","created_at_i":1563631652,"num_comments":0,"objectID":"20485830","points":2,"story_id":20485830,"title":"A Go web development eco-system, designed to life easier","updated_at":"2024-09-20T04:40:15Z","url":"https://gobuffalo.io/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"thunderbong"},"title":{"matchLevel":"none","matchedWords":[],"value":"Buffalo \u2013 Rapid Web Development in Go"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"https://gobuffalo.io/"}},"_tags":["story","author_thunderbong","story_31584418"],"author":"thunderbong","created_at":"2022-06-01T16:35:49Z","created_at_i":1654101349,"num_comments":0,"objectID":"31584418","points":1,"story_id":31584418,"title":"Buffalo \u2013 Rapid Web Development in Go","updated_at":"2024-09-20T11:18:49Z","url":"https://gobuffalo.io/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Liriel"},"title":{"matchLevel":"none","matchedWords":[],"value":"Pkger \u2013 Static File Embedding in Go"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["gopuff"],"value":"https://blog.gobuffalo.io/introducing-pkger-static-file-embedding-in-go-1ce76dc79c65"}},"_tags":["story","author_Liriel","story_21572918"],"author":"Liriel","created_at":"2019-11-19T14:10:17Z","created_at_i":1574172617,"num_comments":0,"objectID":"21572918","points":1,"story_id":21572918,"title":"Pkger \u2013 Static File Embedding in Go","updated_at":"2024-09-20T05:12:13Z","url":"https://blog.gobuffalo.io/introducing-pkger-static-file-embedding-in-go-1ce76dc79c65"}],"hitsPerPage":50,"nbHits":28,"nbPages":1,"page":0,"params":"query=gopuff&tags=story&hitsPerPage=50&advancedSyntax=true&analyticsTags=backend","processingTimeMS":7,"processingTimingsMS":{"_request":{"roundTrip":18},"afterFetch":{"format":{"highlighting":1,"total":1},"merge":{"mergeLoop":{"prepareNextHit":1,"total":1},"total":1},"total":2},"fetch":{"query":2,"scanning":1,"total":4},"total":7},"query":"gopuff","serverTimeMS":8}