Sending a big high-five to Param Jaggi and the team at Agora this week for shipping a 10x search improvement in 3 days (!) with the switch to AI-native search with Objective! Check out how they did it — https://lnkd.in/g4Zi-_6Q
Objective, Inc.
Software Development
San Francisco, CA 973 followers
A modern search platform that learns from human feedback.
About us
A modern search platform that learns from human feedback. Say hi at [email protected]
- Website
-
https://www.objective.inc
External link for Objective, Inc.
- Industry
- Software Development
- Company size
- 11-50 employees
- Headquarters
- San Francisco, CA
- Type
- Privately Held
- Founded
- 2021
- Specialties
- Machine Learning, Content Understanding, and Search
Locations
-
Primary
San Francisco, CA, US
Employees at Objective, Inc.
-
Andres Krogh-Walker
Marketing @ Objective, Inc. All 200s start with 500s.
-
Bruno Miranda
Senior Vice President of Engineering @ Doximity (NYSE:DOCS)
-
Pablo Mendes
Co-founder & CEO @ Objective, Inc | PhD, Artificial Intelligence
-
Ozéias Sant'ana
Engineering Leader | Fractional CTO | Advisor | Startups
Updates
-
Objective, Inc. reposted this
We just shipped a new demo comparing keyword search to AI search. The results are pretty crazy. If you're still using a keyword based search system, I'd strongly recommend looking into how you can apply machine learning to search. We're seeing amazing results with our customers in fashion & e-commerce. If you're interested in how you can easily apply this kind of AI to your own search, don't hesitate to reach out. Try it yourself: https://lnkd.in/d73UMzMC
-
Objective, Inc. reposted this
Ranking is an incredibly important part of search. Search can often get conflated with "relevance", but in practice you need more than that. Whether it's simple sorting or boosting, or more complex ranking like personalization, combining semantic relevance with other Signals like engagement, recency, etc.. is usually required. In Objective we make it easy to leverage your existing data for ranking through Ranking Signals. Simply push the data you want to use for ranking in your objects, and it's available when you're searching to update the order of results. Ranking is fully dynamic, so you can update it on the fly, e.g. to rank different for different use cases or users. I put together a quick how-to showing how to use it, check it out, or read more in the docs: https://lnkd.in/eSsxk3n3
-
Objective, Inc. reposted this
Walmart introduces "Relevance Filtering", a technique for filtering out irrelevant vector search results by predicting a relevance score. While lexical retrieval has a natural stopping point based on keyword overlap, embedding based retrieval lacks this quality. Approaches like top K or thresholding based on cosine similarity fail to generalize across queries. In this paper, they introduce a "Cosine adapter" that uses the query the cosine similarity to map the embedding to an interpretable score that can be used to threshold & filter out results. This approach is similar to Ranked List Refinement, which also uses query dependent features, alongside the embeddings, to filter out irrelevant results. These kinds of approaches lead to great increases in precision, without sacrificing much recall. I'd expect to see more of these kinds of approaches, especially in high precision use cases where there is only 1 correct answer. Paper: https://lnkd.in/esTetv2T Read about Result List Refinement: https://lnkd.in/epN5tBHy
-
-
The team here at Objective HQ is at it again this morning, with Ranking Refinement shipping into Private Beta! Ranking Refinement helps you remove bad vector search results by predicting great, ok, or bad relevance for each result at query-time with a layer of human-like classification, something that vector databases really struggle with.
Remove Bad Results in Vector Search with Ranking Refinement, Now in Private Beta
objective.inc
-
Objective, Inc. reposted this
Your RAG app can only be as good as your retrieval engine. Search quality matters more than ever in the new world. We added a RAG interface to our Wikipedia Search demo to help illustrate. https://lnkd.in/gE_KmgaR
RAG on Objective Wikipedia Search - 5 August 2024
https://www.youtube.com/
-
This morning we’re stoked to ship Auto-Evaluation into your Console! You can’t improve what you don’t measure, especially in search. Auto-Evaluation lets you quantitatively — and comprehensively — measure the relevance of your search. All powered by Objective’s AI-native intelligence layer.
Get Quantitative and Comprehensive Insights into your Search Quality with Auto-Evaluations
objective.inc
-
Want to jump into AI-native search, but not sure where to start? Check out our live demos — they're a great way to get inspiration for your next semantic search project! https://lnkd.in/gq7Cz3wh
-
-
In case you missed our Wikipedia experiment this morning, we built a live side-by-side demo to go along with it. https://lnkd.in/gYM4a69U It's a really great, tactile way to feel the difference that AI-native search gives you. Take it for a spin — search for anything that might be in Wikipedia!
-