Chalk
Chalk is a San Francisco-based AI infrastructure company operating a real-time, in-VPC data platform for ML inference and AI agents, serving enterprise customers across fintech, marketplaces, consumer, healthcare, and security with sub-5ms feature serving and gVisor-isolated agent sandboxes.
- Company typePrivate
- Founded2023
- HeadquartersSan Francisco, United States
- Headcount11–50
- GTM typeB2B
- OfferingSoftware
What Chalk does
Chalk is a San Francisco-based AI infrastructure company founded in 2022 that operates a real-time data platform for machine learning inference and AI agents. The platform centers on the Chalk Context Engine, a temporal feature store that versions every feature by time and serves features with sub-5 millisecond p99 latency at up to 100,000 QPS per engine, eliminating train/serve skew through a single Python source of truth used across training and serving. A Rust-powered runtime transpiles user-defined Python resolvers into native code via the Symbolic Python Interpreter, and Chalk Compute (launched June 2026) extends the context engine with gVisor-isolated, in-VPC agent sandboxes and an MCP Gateway that enforces knowledge-cutoff propagation across tool calls.
The product is sold as a unified platform composed of six modules — Feature Store, Real-Time Serving, Temporal Aggregations, Training Data, LLM Toolchain, and Chalk Compute — packaged into vertical solutions for fraud detection, payments risk, credit underwriting, recommender systems, search and ranking, and growth decisioning. Chalk runs natively inside customer AWS (EKS), GCP (GKE), and (coming soon) Azure (AKS) accounts, aligning procurement with existing cloud commitments and meeting strict data-residency requirements. Pricing is quote-based; the primary go-to-market combines enterprise field sales ('Book Demo' / 'Talk to an Engineer' CTAs) with a developer-led adoption path via public documentation, a Python SDK, CLI, and an open GitHub examples repository.
The company has raised approximately $60 million across a $10M seed (December 2023, General Catalyst-led) and a $50M Series A (May 2025, Felicis Ventures-led at a $500M post-money valuation), giving it a 50x markup in 17 months. It serves 15+ named enterprise customers spanning fintech (Ramp, MoneyLion, Mission Lane, Iwoca, Melio, Pipe, Socure, Persona, Found), marketplaces (Whatnot, Apartment List, Turo), consumer (Grindr, MyFitnessPal), healthcare (Medely, Vital), fraud and security (Verisoul, Doppel, Musubi Labs), workforce management (Nowsta), and residential solar (Sunrun). Chalk was recognized on CB Insights AI 100 in 2024, Fast Company's Next Big Things in Tech 2025, and Fast Company's Most Innovative Data Science Companies 2026.
Chalk firmographics
Firmographics- Name
- Chalk
- Legal name
- Chalk AI Inc.
- Website
- https://chalk.ai
- Company type
- Private
- Founded year
- 2023
- Operating status
- Operating
- Headcount range
- 11–50 employees
- Short description
- Chalk is a San Francisco-based AI infrastructure company operating a real-time, in-VPC data platform for ML inference and AI agents, serving enterprise customers across fintech, marketplaces, consumer, healthcare, and security with sub-5ms feature serving and gVisor-isolated agent sandboxes.
- Ownership category
- akta.pro rank
Where Chalk is headquartered
LocationHeadquarters
- HQ city
- San Francisco
- HQ country
- United States
- HQ region
- North America
Offices2 records
Markets served
Chalk business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Technology or R&D, Personnel, Operations, Marketing or Sales, Infrastructure
Revenue model
- Enterprise SaaS subscription (quote-based): Pricing is not publicly disclosed; customers engage via 'Book Demo' / 'Talk to an Engineer' CTAs and receive customized enterprise contracts. The platform deploys inside the customer's own cloud (AWS/GCP/Azure), suggesting infrastructure-aligned recurring revenue tied to compute and feature volume. Enterprise contracts typically support production ML workloads serving millions of features per second at single-digit millisecond latency, implying significant per-customer contract values.
- Chalk Compute (agent runtime) — incremental add-on revenue: Chalk Compute launched June 2026 as an enterprise runtime for AI agents and inference, available to existing customers running on AWS EKS and GCP GKE with Azure AKS coming soon. Expected to expand ACV via agent sandbox and workload execution volume.
- Professional services / Forward Deployed Engineering (implied): Customer testimonials repeatedly reference 'the Chalk team' working closely with customers on infrastructure tuning, latency tail issues, scaling readiness, and custom feature builds (e.g., Verisoul, Apartment List). Forward-deployed engineers are listed in blog author bylines (Melanie Chen - FDE Manager), indicating professional services attached to enterprise contracts.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Other | Annual | Custom enterprise quote — not publicly listed |
Go-to-market motion3 records
Distribution channels3 records
Marketing channels8 records
Chalk product offering
Product offeringCore offering
Chalk is a real-time AI data platform that provides the infrastructure needed to operationalize machine learning and AI inference. The platform combines a temporal feature store, real-time serving engine, training data generation, temporal aggregations, and an LLM toolchain into a unified Python-first execution layer, deployed inside the customer's own AWS, GCP, or Azure cloud account. Chalk Compute extends the platform with isolated agent sandboxes for time-traveling agent evaluations and autonomous workloads.
Product overview
Chalk is a single AI data platform sold as a unified product composed of multiple integrated modules, all running inside the customer's own AWS, GCP, or Azure cloud account. The platform centers on the Chalk Context Engine, which combines a feature store, real-time serving, temporal aggregations, training data generation, and an LLM toolchain into one Python-first, compute-first execution layer. Chalk Compute is the newest module, extending the same context engine with isolated agent sandboxes for time-traveling evaluations and autonomous workloads. Vertical solutions for fraud detection, payments, credit underwriting, recommender systems, search and ranking, and growth decisioning package these capabilities into use-case-specific offerings.
Differentiator
Problem solved
Functional benefit
Brands
- Chalk Compute: Enterprise-grade agent runtime that deploys sandboxes directly inside the customer's private cloud, integrated with the Chalk Context Engine.
- Chalk Context Engine
Products and services
- Chalk Feature Store
- Chalk Real-Time Serving
- Chalk Temporal Aggregations
- Chalk Training Data
- Chalk LLM Toolchain
- Chalk Compute
Companies that use Chalk
Customer profileNamed customers21 records
Segments8 records
Ideal customer profiles4 records
Chalk technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration17 records
AI capability15 records
Feature9 records
Chalk partnerships and signals
Strategic signalPartnerships
Eleven partnerships are on record, tiered strategic cloud partner, emerging cloud partner, strategic data warehouse partner, standard data source integration, operational tooling partner, orchestration integration and agent-harness ecosystem.
- Amazon Web Services (AWS)strategic cloud partnerChalk runs natively inside customer AWS accounts (EKS-based deployment for Chalk Compute). Documented integration with AWS services and used by Whatnot, MoneyLion, and Turo on AWS.
- Google Cloud Platform (GCP)strategic cloud partnerChalk Compute deploys to GCP via GKE; documented GCP integration on chalk.ai and used by Apartment List on GCP.
- Microsoft Azureemerging cloud partnerChalk Compute support for Azure (AKS) announced as 'coming soon' in June 2026. SSO sign-in available via Azure Active Directory.
- Snowflakestrategic data warehouse partnerDocumented integration with Snowflake; Chalk's Context Engine and feature store natively connect to Snowflake as both a query source and online/offline store. Listed as a Chalk integration on the homepage.
- Databricksstrategic data warehouse partnerDocumented Databricks integration on chalk.ai homepage; Chalk positions itself as a Databricks competitor in the news but also offers native integration with Databricks as a data source.
- PostgreSQLstandard data source integrationDocumented PostgreSQL integration allowing customers to use their existing Postgres database as both online and offline feature store with no bespoke storage.
- Datadogoperational tooling partnerListed integration on chalk.ai homepage for observability and monitoring of ML workloads.
- PagerDutyoperational tooling partnerListed integration on chalk.ai homepage for incident alerting on production ML pipelines.
- Slackoperational tooling partnerListed integration on chalk.ai homepage for notifications and team collaboration on ML workflows.
- Apache Airfloworchestration integrationDocumented Airflow integration with a public example repository showing how Chalk resolvers can be triggered from Airflow DAGs.
- OpenAI / Anthropic / LangChain / LlamaIndex / MCP serversagent-harness ecosystemChalk Compute is designed to be harness-agnostic; supports the OpenAI Agent SDK, Anthropic SDK, LangChain, LlamaIndex, and arbitrary MCP servers. Customers can bring their own agent framework and run it on Chalk infrastructure.
Scale indicators10 records
Recent moves8 records
Expansion highlights6 records
Chalk competitors and assessment
Company assessmentDirect peers
- Tecton: Direct feature-store competitor serving enterprise ML teams with real-time feature serving and online/offline consistency. Most directly comparable to Chalk's feature store and real-time serving modules across fintech and customer personalization workloads.
- Hopsworks: Feature store platform with online/offline parity and Python-first design, competing for the same ML platform buyer personas. Comparable in target market (enterprise ML teams) and deployment model (cloud-native).
- Modal: Cloud compute platform for AI workloads and agents. Chalk explicitly cites Modal as a generic agent runtime Chalk Compute is differentiated against; directly comparable in the agent runtime/inference category.
- E2B: Open-source agent sandbox runtime for AI code execution. Chalk's gVisor-based in-VPC sandboxes are positioned against E2B's hosted model, making them a direct competitor in the agent compute category.
- Daytona: Development environment and AI agent infrastructure platform. Cited alongside Modal and E2B as a comparable agent-runtime competitor that Chalk Compute differentiates against through in-VPC and temporal consistency.
Broad incumbents
- Databricks: Broad data and AI platform with a bundled Feature Store and MosaicML/Agent Bricks offerings. Reuters and Chalk's own materials frame the two as direct competitors in AI inference and feature serving for enterprise ML.
- AWS SageMaker: Hyperscaler ML platform with bundled Feature Store, inference, and MLOps capabilities. Chalk's in-VPC AWS deployment puts it inside the buyer's AWS account, making SageMaker the primary bundled incumbent competing for the same ML platform budget.
- Google Vertex AI: Hyperscaler ML/AI platform with a Feature Store and Agents offering. Directly competes with Chalk on GCP deployments (Apartment List runs on GCP) and is a co-sell partner and competitor simultaneously.
- Weights & Biases: MLOps and experiment tracking platform that overlaps with Chalk's ML development workflows, training data generation, and observability features. Comparable enterprise buyer (ML engineering leaders) with broader ML lifecycle coverage.
Emerging players
- Feast: Open-source feature store that overlaps with Chalk's training-data and online feature serving for ML teams. Differs in being OSS-first vs Chalk's commercial enterprise focus; pressures Chalk on open-source adoption paths.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks6 records
Key highlights7 records
Customer concentration
Chalk social profiles
Digital presenceChalk compliance and trust
Trust signalCompliance3 records
Chalk financial estimates
Financial estimateRevenue estimate
Valuation estimate
Chalk leadership team
Management profileNumber of profiles
Profiles9 records
Chalk funding detail
Funding detailFunding overview
Funding rounds5 records
Investors10 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Chalk M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Chalk
What does Chalk do?
Chalk is a real-time AI data platform that provides the infrastructure needed to operationalize machine learning and AI inference. The platform combines a temporal feature store, real-time serving engine, training data generation, temporal aggregations, and an LLM toolchain into a unified Python-first execution layer, deployed inside the customer's own AWS, GCP, or Azure cloud account. Chalk Compute extends the platform with isolated agent sandboxes for time-traveling agent evaluations and autonomous workloads.
Is Chalk a public or private company?
Chalk is a private company. It is classified as venture growth investor backed and is currently operating.
When was Chalk founded?
Chalk was founded in 2023. It employs 11 to 50 people.
Where is Chalk based?
Chalk is headquartered in San Francisco, United States, in the North America region.
How does Chalk make money?
Three revenue lines are on record. Enterprise SaaS subscription (quote-based) is the primary driver. The others are chalk Compute (agent runtime) — incremental add-on revenue and professional services / Forward Deployed Engineering (implied).
Who are Chalk's main competitors?
Direct peers on record are Tecton, Hopsworks, Modal, E2B and Daytona. Broad incumbents are Databricks, AWS SageMaker, Google Vertex AI and Weights & Biases. Feast is listed as an emerging player.
Does Chalk have an API?
Yes. Chalk offers a developer-facing API surface through its Python client library (ChalkClient) used to query online features, run offline queries, and manage deployments. Code examples include `ChalkClient().query({...})` for synchronous online feature retrieval and `client.offline_query({...})` for distributed offline training data generation. A CLI is also provided. Authentication supports Google and Azure Active Directory SSO, plus per-account email sign-in. Rate limits, versioning, and sandbox tiers are not explicitly detailed in the source. Developer documentation is at docs.chalk.ai/api-docs.