# Chalk

> Chalk builds the data and compute layer that supplies fresh features to machine-learning models at the moment a decision is made. Its Python-defined feature platform unifies training, batch scoring, and live inference; newer products add model serving, agent sandboxes, and production-connected notebooks. Customers use it for risk, recommendations, pricing, and other time-sensitive decisions.

- **Founded:** 2022
- **Headquarters:** San Francisco, United States
- **Founders:** Marc Freed-Finnegan (Co-founder and CEO), Elliot Marx (Co-founder), Andy Moreland (Co-founder)
- **Products:** Chalk Context and feature store, Model serving and model gateway, Chalk Compute, Chalk MCP Server and Chalk Assistant, Chalk Notebooks
- **Notable:** Raised $60 million across a $10 million seed and $50 million Series A by May 2025., Named to CB Insights' 2024 AI 100 list., Whatnot reports 300 million-plus features served per second and P99 feature latency below 100 milliseconds on its workload.

## Products & services

- **Chalk Context and feature store** — Python-defined features and a query engine that serve precomputed or on-demand values consistently across training, batch scoring, and live inference.
- **Model serving and model gateway** — Runtime products for deploying models and routing inference requests alongside Chalk's feature infrastructure.
- **Chalk Compute** — Agent sandboxes in a customer's cloud, integrated with point-in-time context and evaluation workflows.
- **Chalk MCP Server and Chalk Assistant** — Agent access to Chalk queries, feature definitions, and ML workflows, plus an assistant within the product.
- **Chalk Notebooks** — Production-connected Python and SQL notebooks that can query branches and point-in-time data inside a customer's cloud.
- **Forward-deployed engineering** — Hands-on engineering support for production integrations and scaling.

## Achievements

- Raised $60 million across a $10 million seed and $50 million Series A by May 2025.
- Named to CB Insights' 2024 AI 100 list.
- Whatnot reports 300 million-plus features served per second and P99 feature latency below 100 milliseconds on its workload.
- Whatnot reports expanding real-time personalization coverage from about 90% to 99.9%.

## Latest updates

- **2026-09** — Launched Chalk Notebooks for production-connected, point-in-time ML analysis.
- **2026-09** — Announced Chalk MCP Server and Chalk Assistant for agent-supported ML workflows.
- **2026-06** — Announced Chalk Compute, an in-cloud runtime for agent sandboxes and historical evaluation.
- **2026-02** — Added dashboard webhooks, vector-database registration, and runtime performance improvements.

## Links

- Website: https://chalk.ai
- LinkedIn: https://www.linkedin.com/company/chalkai
- Twitter/X: https://x.com/chalk_ai
- GitHub: https://github.com/chalk-ai

---

Profile page: https://yespress.io/chalk
Published by YesPress — https://yespress.io
Last updated: 2026-09-28