Suggestions
Talip Ozturk
Founder of A Vector Database Startup
Talip Ozturk is an entrepreneur and technology leader based in the San Francisco Bay Area, currently serving as the Founder of a stealth startup focused on vector databases. He previously held the position of Head of Engineering at Vectara, an AI-powered search platform that utilizes neural search technology to enhance user experiences in information retrieval.
Professional Background
-
Vectara: Talip was instrumental in launching Vectara, which aims to simplify search solutions by eliminating the complexities often associated with traditional search engines like Elastic Search and Solr. His work there involved leveraging natural language processing (NLP) to improve search functionality without extensive configuration.12
-
Education: Talip studied Mathematics at Middle East Technical University but did not complete his degree. He later earned a Bachelor's degree in International Business from Johnson & Wales University, where he attended from 1997 to 2000.2
Current Endeavors
Since leaving Vectara, Talip has embarked on a new venture focused on developing a vector database, indicating his continued commitment to advancing technology in the search and data management space.345 His expertise in engineering and product development positions him as a significant player in this emerging field.
Highlights
About 6 months ago, I left Vectara, an AI startup, for a new beginning. I learned so much from the amazing people at Vectara. Thank you Amr Awadallah, Amin Ahmad, and Tallat M. Shafaat for the opportunity. While my friends at Vectara are providing the best end-to-end RAG solution as a service, I decided to focus on a very specific problem that is closer to my heart, my past.
I believe that eventually, AI will become an essential part of every software product. Just like SQL today, we as developers will become adept and comfortable working with AI models and vectors. We will start seeing projects working with billions of vectors. A high number of vectors result in very high latency and lower recall. Pick a very famous vector database and insert 1 billion vectors into a single index/namespace and check the query latency yourself. Painful and disappointing experience…
At Vectara, I saw firsthand how challenging it is to scale vectors. After Vectara, I focused on building a solution to achieve single-digit ms. latency for a billion-vector index. Just like the early days of Hazelcast, for months, I have been shuffling between “this is too difficult to do” and “I can do it”. More on this later.
For this journey, I am looking for design partners to shape the product. If you have vector database challenges, if scale and/or latency are a big deal for you, I would love to talk to you. Please reach out.
#billion_vectors #single_digit_latency #no_joke
Very happy to announce our state of the art information retrieval model, Boomerang! Vectara is an end to end API-driven LLM based search service so turning your text/documents into vector embeddings with Boomerang is one of the many awesome things Vectara does for you behind the scenes. We have more things cooking... stay tuned.
