Subscribe
Sign in
Tilores Identity RAG ā Customer data search, unification and retrieval for LLMs
Data scientists connect Tilores to their LLM to search internal customer data scattered across multiple source systems. The LLM retrieves unified customer data, which it uses to answer queries or as context when querying subsequent unstructured data.
76
Replies
Best
@major_grooves Interesting. So in my app, we crawl travel data from across hundreds of travel sites, articles and feed those data packets onto LLM based on user's query. Is there anyway Tilores can assist in the entire process? Super congratulations on the launch.
@akshay_lahri how do you currently feed that data into the LLM? A vector database? If you end up with lots of duplicate records when you are crawling, Tilores might be able to help you deduplicate them, but tbh when it comes to the unstuctured text in a typical website article, you might be best sticking with a vector database.
@major_grooves congrats, this is really useful. We have a lot of duplicative records in our CRM, especially when folks move on from one co to another, this could be a great extra layer to remove one frequent failure mode.
@major_grooves congratulations on the launch. Cool product! How does Tilores RAG deal with data versioning? So how does it track the change in data over time and can I perform a search based on a specific point in time? Cheers
Congratulations on the launch. It looks really cool and powerful, and it seems like something that will simplify many use cases. I'm looking forward to seeing what people build with it.
Thank you @janoberhauser. I opens the door for so many different use cases and allows all Python users to easily access their data in Tilores. I am looking forward to see people using it from within N8N.
We will run a webinar next week to show how easy it is to start using Tilores IdentityRAG together with LangChain: https://www.linkedin.com/events/...
How nice is that! That seems like a an amazing piece of technology. Truly innovative! Will test it soon!!!
@mathias_n we can even get you set up and running in person in Berlin!
Nice, will recommend it to my marketing agency. I love the RAG feature.
Does it also support real time syncing my contacts across LinkedIn, Google Sheets, Eventbrite, Meetup and HubSpot in one database?
@arthurpoot yes - as long as we can access them via API we can both pull data from them and push data back to keep them updated - in real-time. Glad you like it!
@nicolas39 so Tilores is designed to be highly-scaling with zero input required, since we use serverless technology. So we can ingest as much data and provide as many searches, in parallel, as you could ever need. The only thing you would have to keep an eye on is the cost of the LLMs themselves, since that is outwith our control.
Steven, as a data scientist constantly battling fragmented customer info, Tilores Identity RAG feels like the missing puzzle piece I didn't know I needed. Your solution's ability to seamlessly integrate with multiple data sources and create that unified 'source of truth' is not just technologically impressiveāit's a relief for those of us striving for hyper-personalized customer experiences.
The fact that Tilores can keep up with real-time updates and is tailored for European data privacy standards is a huge plus. It's fantastic to see a tool that not only streamlines data handling but also respects the intricate web of GDPR compliance. Kudos on that front!
One question that pops up is how flexible Tilores is in adapting to different LLM frameworks beyond LangChain? And for those of us diving deep, are there any plans to introduce advanced analytics or visualizations to help us better understand customer patterns?
Excited to give it a spin and see how it elevates our LLM's performance. That $500 credit offer is a generous nudge to kickstart the journey. Heading over to the GitHub repo now!
@frank_petron we can also work with Bedrock, which es effectively AWS's framework. As for other frameworks, we would certainly integrate with more if there is demand.
Let us know how you get on with the integration! We are happy to jump onto a call to help you 1:1.
It's so amazing! I'm thinking about how to integrate RAG into my website https://quitporn.ai. And today I found your tool. It's a perfect example for me to learn and use. Congrats to your team!
Hey Steven,
How quickly can the system update and retrieve unified customer profiles?
Have you considered expanding beyond customer data to other domains where entity resolution could be valuable?
Congrats on the launch!
Hi @kyrylosilin the process to update a profile takes less than 500 ms. - No matter how many profiles change at the same time.
We are also using Tilores in other spaces like company data.
Thank you
Hendrik
@kyrylosilin I found their API to be REALLY fast, even at scale. And what I found really nice comparing to other entity resolution systems, you can define the golden record at read-time, as opposed to at write time. This way you can get different perspectives on the same source data.
For example, for some applications you might want to have the latest email of an unified customer.
For some applications you might want to have all emails that belong to one unified customer.
The ability to define different golden records at read-time makes Tilores really flexible.
Congrats to the team on this milestone! Excited to see how Tilores is evolving!
@irhymeth your support and the support of the good data people, has been invaluable!
Launching soon!
congratulations on your launch @major_grooves. How does Tilores ensure the accuracy and consistency of the unified customer data when pulling from multiple sources?
@major_grooves @zishaniqbal the data is normalized and transformed before matching. Also certain attributes can get prioritized. Let us jump onto a call if you have further questions.
@zishaniqbal good question. We have a pretty sophisticated, predominantly rules-based matching engine that uses various fuzzy matching algorithms, such as Cosine similarity, to do matching on strings, such as name. We also have ML based matching. The accuracy of the matching depends on how well the rules are set - but the important thing is that the rules that are used for matching are also mentioned in the identity graph, so they are highly explainable.
Congratulations Tilores! Data scientists can now seamlessly link LLMs to scattered internal customer data, ensuring smarter queries and better results. This is a game-changer! š” #AIInnovation #DataIntegration
wow @major_grooves, hats off.
love the idea of seamlessly unifying scattered data - sounds like a game-changer.
Congrats to the Tilores team on the launch of Identity RAG! This sounds like a powerful tool for streamlining access to unified customer data. Are there any specific integrations available for different source systems to enhance data retrieval?
@vietpham there is a snowflake and webhook integration, howeverwe also provide a graphql API and a python sdk which easily integrates with most systems
@vietpham our most used integrator is actually for Snowflake. Other than that, using our GraphQL API you can connect to any data source or we can discuss building a specific connector for you for specific sources.
š Congratulations on the launch of Tilores Identity RAG! š
Iāve created an interactive demo of your product using Supademo to showcase its features in action. You can check it out here: https://app.supademo.com/demo/cm...
Supademo allows you to create engaging, interactive demos in minutes, and it's a great way to highlight key aspects of your product for potential users. I'd love to see this demo added to your launch pageāit could help drive more engagement during your launch and beyond!
Feel free to share it with your community and let us know if youād like any updates or adjustments.
Best of luck with your launch!
Thank you @vimal_kumar_rai that is a nice tool that you created there.
Tilores