We spent last week at Snowflake Summit. The question we heard most was: How do we trust agents in production? Teams building agents today are evaluating them against synthetic test environments, making educated guesses about how they'll behave, and shipping with incomplete confidence. When something goes wrong, reconstructing what the agent saw and why it decided what it did is either painful or impossible. The bottleneck has shifted from model quality to infrastructure quality. Grindr showed conferencegoers how it approaches the agentic infrastructure supporting 15 million users (hint: having Chalk helps). Chalk co-founder Elliot Marx demoed what solving the infra bottleneck actually looks like in practice to a standing-room-only crowd with our latest product: an enterprise-grade agent runtime that deploys sandboxes directly inside your private cloud. We call it Chalk Compute. Get up to speed. Link in the comments. #SnowflakeSummit #Chalk
Chalk
Software Development
San Francisco, California 10,491 followers
The AI data platform.
About us
Context+ infra for agents + models. Built for speed. In your cloud.
- Website
-
https://chalk.ai?utm=linkedin
External link for Chalk
- Industry
- Software Development
- Company size
- 51-200 employees
- Headquarters
- San Francisco, California
- Type
- Privately Held
Products
![Click here to view Chalk]()
Chalk
Data Science & Machine Learning Platforms
Chalk is a data platform that powers machine learning and generative AI. Chalk’s best-in-class developer experience enables data teams to declare features and their dependencies with idiomatic Python in online, streaming, and batch environments. Chalk compiles these definitions into parallel pipelines that run on a Rust-based engine. These pipelines use the exact same source code to serve temporally-consistent training sets to data scientists and live feature values to models. This re-use ensures that feature values from online and offline contexts match and dramatically cuts development time. With Chalk, engineers, data scientists, and analysts can focus on their unique products while Chalk seamlessly handles data infrastructure.
Employees at Chalk
View 119 employees at Chalk
or
Locations
-
Primary
Get directions
2390 Mission St
San Francisco, California 94110, US
-
Get directions
54 W 21st St
590
New York, NY 10010, US
Updates
-
Chalk reposted this
chalk's first big strategic event is in the books. up until now, our growth has come from one of the best SDR teams in data and a stream of organic inbound and word of mouth. Snowflake summit was a different motion. and this team showed up. a few things i'll be thinking about for a while: Marc and AJ Balance (CPO at Grindr) packed the room on the ML track. people were still messaging me days later bummed they couldn't get in. chalk compute went live publicly right after. grindr endorsed it from stage. one of those moments you plan for and still can't quite believe when it happens. Elliot followed with a 20-minute deep dive on chalk inside the snowflake stack. hundreds of people showed up. exactly the audience we've been building for. developers are wired to find hard problems. so we put a real math proof on a chalkboard at the booth and gave crewnecks to anyone who solved it. proofs got harder each day. we still ran out of winners. Stephen, a month into chalk, told me on day 3 it didn't look like our first conference. highest compliment a scaling team can get. a few specific shoutouts: Lina, who jumped into field marketing two weeks before this whole thing and made it look easy David Quinn and Fahim Zaman, both weeks into the role, ran the booth, the messaging, and the launch all together Hyun Auh, who designed every piece of it Chris Studley, who pointed the SDR team at summit and kept the meetings flowing all four days the FDEs, who demoed compute on the floor on day one of its public life most of this team has been at chalk under 90 days. they pulled this off. i'm still kind of stunned. onto next week. something else is a brewin. watch chalk.ai/speed
-
-
-
-
-
+1
-
-
Two new faces this week. Shawn Wang joins as an Account Executive, coming from Solo where he worked with customers on Kubernetes and cloud infrastructure. He lives in San Francisco with his wife and spends his free time exploring the city. He grew up split between North Dakota and Hong Kong, which made for an interesting childhood. Leonhard Spiegelberg joins as an Engineering Manager, bringing experience from OpenAI, Snowflake, and BMW, plus five and a half years of database research at Brown. He has strong opinions about pasta versus potatoes (pasta wins, except for fries), once spent 1.5 years optimizing CSV parsing, and prefers his 1:1s on a walk. Welcome to both. We're glad you're here. #Chalk #Hiring
-
-
Building agents is easy. Trusting them in production is not. We put a bus on the streets of San Francisco this week asking: What do you do when latency > 5ms means it blows up? It's the actual constraint our customers operate under every day. The bottleneck is the outer loop: evaluating against real historical context, fixing what broke, and redeploying with confidence. Most runtimes can't do that because they don't own the data layer. Chalk Compute does. Lock an agent to any point in time and run it against the exact data your production system would have served at that moment. Inside your VPC. No synthetic test sets. Real evals. Get up to speed. Link in the comments. #SnowflakeSummit #Chalk
-
-
Chalk reposted this
few things hit like getting sent a video of Candice-Lee James chasing the chalk speed bus down the street outside snowflake summit. i'm so proud of this team. we painted a bus to launch time-traveling agent sandboxes. went back to 1994 when Keanu was king and AI was science fiction to ship the future. sounds like us. built for teams that need to go fast. in your cloud
-
More math at booth #2611 at Snowflake Summit Day 2. Today's problem: the Gaussian Integral. Today's problem setter: John Rotter, FDE at Chalk and former experimental high energy particle physicist. His PhD thesis measured the lifetime of the Higgs Boson using off-shell H to WW events in the Vector Boson Fusion channel. Translation: he is exactly the person you want writing your math problems. Drop by the booth to take a shot. Correct answers win a Chalk Lululemon sweatshirt, an item with no official market price and a strong secondary one.
-
-
Pop quiz, hotshot. Latency > 5ms = it blows up. What do you do? We put that on a bus. Because the teams building real-time AI (MoneyLion, Whatnot, Grindr, Socure, Turo) don't get to round up on latency or on trust. They need agents that run on real data, inside their own cloud, with the confidence that what worked in testing will hold in production. Yesterday we shipped Chalk Compute. Agent sandboxes that run in your cloud, with time-traveling evals so your agents test against the exact data they would have seen in production. No synthetic environments. No data leaving your VPC. Get up to speed. Link in the comments. #SnowflakeSummit #Chalk
-
-
Elliot is on stage at Snowflake Summit at 1pm PDT today, making the case for a computation-first architecture for real-time feature stores. But real-time feature stores are only half the puzzle. The other half is Compute, which we launched yesterday. He'll walk through both with live code. Find Elliot and our engineers at booth #2611 after for a deeper dive. Basecamp South Theater 2 (South Hall A). Session ML202.
-