🗓️ Hungry for even more AI events? Check out SF IRL, MLOps SF, or Cerebral Valley’s spreadsheet!

🚀 Don’t miss the UoT chapter launch! Geoffrey Hinton will be speaking at UofT from 4-5:30pm, followed by a 6-8pm launch and event mixer. A report from the Globe and Mail will be onsite to talk with attendees about AI, Hinton, and more.

This week we’re excited to share our recent interview with our Head of Media and Marketing and the founder of the Collective Intelligence AI Podcast, Thomas Joshi and the AI Product Lead at Timescale, Avthar Sewrathan.

Timescale has built a Postgres database that is revolutionizing AI application development by seamlessly integrating vector search capabilities into traditional relational databases. Thomas and Avthar discuss how their innovative approach enables developers to build powerful RAG systems and AI agents without the complexity of managing separate vector databases.

Thomas (GenAI Collective): “Welcome, Collective Nation, to another installment of the GenAI Collective podcast, Collective Intelligence. We have an AI lead at Timescale with us. Do you mind explaining what Timescale is and your role there?”

Avthar (Timescale): “Fantastic, Thomas, thanks so much for having me. My name is Avthar. I lead AI products at Timescale. We're a Postgres database company that’s been around for nearly 10 years, helping developers build applications over and above general relational database use cases. We focus on time series and real-time analytics, and my product group is all about AI and vector data. We build products enabling developers to create AI systems like RAG, semantic search, and even AI agents using Postgres as the core database.

“Our core products include PG AI, which complements the open-source extension PGVector. All our work is open source. I’m really excited to chat about how we’re helping developers build AI applications.”

Thomas: “What does your role look like day to day?”

Avthar: “I’m the product lead, or PM, focusing on how to build AI applications that give reliable, accurate answers. Retrieval-augmented generation (RAG) is often the solution. We also look at structured data retrieval, not just unstructured text or PDFs. Our job is to enable developers to build useful applications with minimal headache, using the tools they already know—extending Postgres for AI applications. I spend a lot of time talking to developers, listening to their problems, and working with our world-class engineers to help them build the next generation of AI solutions.”

Thomas: “Let’s walk through the problems people face when moving from relational databases to non-relational databases—and why they come to Timescale.”

Avthar: “When ChatGPT arrived, everyone realized, ‘I want ChatGPT’s power, but trained on my business or private data.’ That need brought about vector databases, which allow you to search unstructured data—texts, PDFs—by creating vector representations. It’s how you get the classic ‘chat with these documents’ use case, where you search a knowledge base for the most relevant pieces of text, feed them into the model’s context, and get targeted answers.

“The question became: do you need a separate vector database? That means more infrastructure, a new query language, syncing data across two systems. A lot of developers asked, ‘Is there a way to get vector search without adding a separate database?’ PGVector effectively adds vector functionality onto Postgres. Timescale’s PGVector Scale extension boosts performance, and PG AI automates embedding creation. Our philosophy is: use the Postgres you already know and love, eliminate extra complexity, and build robust AI applications.

“PGVector provides the core storage and search for vectors, while our additions make it easier to handle real-world demands like 50 million or even a billion embeddings. We also handle advanced filters—multi-tenant data, time-based constraints. All these pieces are crucial to building AI apps that give accurate answers. We solve these problems so developers can move faster, with less friction.”

Thomas: “There’s a spectrum of solutions. What’s the sweet spot customer persona for Timescale?”

Avthar: “You can’t build for everyone. Our typical customer has large-scale RAG or agent applications. For instance, the financial industry: analyzing massive amounts of research, PDFs, and real-time data. One of our customers, Market Reader, offers real-time market insights. They track news, integrate that with stock data, and deliver summaries to users—all backed by Postgres and Timescale.

“We also excel at multi-tenant or more complex scenarios that need filtering and joins on business data. If you’re building a user-facing SaaS, you likely have different permission tiers and want to apply time-based or user-based filters. Handling it inside a single database, with simple SQL joins, is far simpler than syncing multiple systems.

“A lot of specialized vector solutions evolved from narrower tasks, like Facebook AI Similarity Search (Faiss), which is purely about high-performance vector retrieval. But real-world RAG apps often demand joins, permissions, user data, and more. Postgres is naturally strong at that, and we add vector and AI functionality on top.”

Thomas: “How do you handle complexity, embedding updates, and developer onboarding?”

Avthar: “Our Vectorizer feature addresses the complexity of embedding pipelines. Typically, you’d have to manually embed new or updated documents and synchronize them in your vector database. We treat it more like a database index: add or delete documents, and the index updates automatically.

“This also makes it easy to experiment with different embedding models. Developers can run quick tests to see if open-source embeddings perform better than, say, OpenAI embeddings. Everything is transparent, but we provide smart defaults. Experienced developers appreciate having the ability to fine-tune and see under the hood, rather than being locked into a black box.”

Thomas: “What are the killer use cases people can dive into right away?”

Avthar: “AI is so universally applicable that we see new use cases daily. A big one is internal co-pilots, especially for industries like energy and IoT, where technicians need fast access to detailed information about machines or updates. Another emerging area involves structured data queries—going beyond unstructured text search to marry it with relational data.

“We see more advanced RAG applications combining PDFs with business data. For instance, e-commerce or finance apps that want product or account information plus relevant documents. As models improve—cost per token drops and reasoning power increases—these applications will multiply. We’re seeing innovative expansions beyond ‘chat with 10 PDFs’ toward real enterprise-level intelligence.”

Thomas: “Last question: how do you see the balance between letting the model handle all reasoning vs. offloading some to the database?”

Avthar: “I think advanced models—like GPT-4, Claude, or DeepSeek—will leverage tools that revolve around a robust database. Even if you have a huge context window, you still need to filter and provide high-quality context. There’s a famous ‘needle in the haystack’ issue where large context windows can confuse models. So the database’s job is less about storing the intelligence itself and more about surfacing the right context—unstructured or structured data—so the LLM can deliver accurate, useful answers.

“Databases, especially Postgres, have been around a long time. They’re ‘Lindy,’ as Nassim Taleb would say, meaning something that’s stood the test of time tends to remain valuable. Combining stable ‘old’ tech like Postgres with cutting-edge AI models and agent architectures is a powerful way to build. We provide that foundation so developers can focus on delivering the best AI experiences.”

Try Timescale’s new PgAI—a PostgreSQL extension for building RAG, semantic search, and AI apps directly in your database. Simplify development and launch faster → Start free

What if you could get introductions to anybody that you wanted to meet?

Meet Boardy - an AI networking partner who makes introductions to investors, founders, recruiters, executives, operators and anyone else that you would like to meet.

No forms, fees or signups required. It’s just a phone call. All you do is 👇🏻

Boardy recently raised an $8M seed round from VC firm Creandum - the guys behind Spotify, Klarna, Depop - and built a network of over 25,000 incredible individuals to connect with.

In the past 6 months alone, Boardy has already:

You’re always one introduction away from changing your life and your business in ways you can’t imagine…

Send Boardy a DM and see what he can do for you.

At the DC Tech Roundtable, policy and tech leaders engaged in a lively discussion on the future of U.S. AI leadership. Conversations centered on leveraging American values—ethically sourced data, integrity-driven innovation, and a legal framework that fosters trust. Participants explored enterprise AI hubs and the role of “little tech” in shaping policy.

The discussion also focused on how to bridge gaps in understanding between industry and government. Attendees left with a sharper sense of how to frame AI’s impact in ways that resonate with policymakers, ensuring innovation and competitiveness remain priorities in the national strategy!

We are a volunteer, non-profit organization – all proceeds solely fund future efforts for the benefit of this incredible community!

About Eric Fett

Eric leads the development of the newsletter and online presence. He is currently an investor at NGP Capital where he focuses on Series A/B investments across enterprise AI, cybersecurity, and industrial technology. He’s passionate about working with early-stage visionaries on their quest to create a better future. When not working, you can find him on a soccer field or at a sushi bar! 🍣

About Noah Frank

Noah is the co-founder of Aurix and has spent his career both working at startups and advising global leaders on innovation strategy. His work and body of research focus on AI policy, anticipatory governance, and effective decision-making. When not working to make emerging tech work for all, you can find him making music with his band.