# Delphina

> Delphina is the AI-managed context layer for messy enterprise data. It captures, validates, and maintains the business context your best analysts carry in their heads, then grounds every AI agent that touches your data in that foundation — so teams get data agents, workflows, and data apps they can actually trust.

AI agents are only as good as the context they have. Plug Claude, Cursor, or any MCP-compatible agent into your warehouse and the answers come back in seconds — and most of them are confidently wrong. The fix isn't a smarter model; it's giving agents the context your best analysts carry in their heads.

Delphina solves this with a context layer that **builds itself, validates itself, and flags what it doesn't know**. Four phases: *Ingest* a living map of your data from warehouse to dashboards, *Refine* it into your business's own definitions and metrics, *Validate* with auto-generated evals plus a built-in critic agent, and *Evolve* with every interaction as schemas and priorities shift. Best-in-class data agents, end-to-end workflows, and shareable data apps then ship on top of that validated foundation.

This is the alternative to the "data context trap" — the failure mode teams hit when they try to build their own context layer (collection, validation, and maintenance compound, and most teams give up after a quarter; see [The data context trap](https://delphina.ai/blog/the-data-context-trap)).

Founded by [Duncan Gilchrist](https://www.linkedin.com/in/dsgilchrist/) (former Director of Data Science at Uber, VP Data Science & Engineering at Gopuff; PhD Harvard) and [Jeremy Hermann](https://www.linkedin.com/in/jeremyhermann/) (former Head of ML Platform at Uber and architect of Michelangelo, co-founder of Tecton). Backed by Costanoa Ventures, Radical Ventures, and 20+ angel investors including Fei-Fei Li, Guido Imbens, Lukas Biewald, and George Sivulka.

## The Foundation: AI-managed context layer

- **Ingest** — Reads your warehouse, dashboards, dbt models, query logs, and existing docs to build a living map of your data landscape
- **Refine** — Systematizes how your business actually works: your definitions, your metrics, the unwritten rules your best analysts carry in their heads
- **Validate** — Auto-generates evals from your trusted dashboards and runs them continuously against live data; a built-in critic agent pressure-tests every answer
- **Evolve** — Learns with every interaction; adapts as schemas update and the team asks new questions
- Full SQL lineage and org-wide observability — nothing is hidden
- MCP-compatible: works with Claude, ChatGPT, Cursor, and custom agents

## The Products (built on the Foundation)

- **Analytics Agent** — Natural-language answers grounded in your context layer; returns SQL, tables, charts, and written analysis, all reviewed by the critic agent
- **Deep Research** — Multi-step research across your warehouse, the kind that used to take an analyst days
- **Workflows** — End-to-end product, marketing, sales, and finance workflows shipped in the box
- **Data Apps** — Shareable apps your whole team can use, grounded in the same validated knowledge

## Deployment & security

- Hosting: secure AWS cloud or single-tenant VPC deployment inside your perimeter
- SOC 2 Type II certified, HIPAA compliant
- Customer data is isolated, authorized, and encrypted
- Your data is never used to train shared models

## Pages

- [Homepage](https://delphina.ai/): The AI-managed context layer for messy enterprise data
- [Our Story](https://delphina.ai/our-story): Team, mission, and the five behaviors that drive how we work
- [Integrations](https://delphina.ai/integrations): Warehouses, BI tools, dbt, and MCP-compatible connectors
- [Security](https://delphina.ai/security): SOC 2 Type II + HIPAA; isolated, authorized, encrypted
- [Docs](https://docs.delphina.ai): API reference, integration guides, and product documentation
- [Book a Demo](https://delphina.ai/book-a-demo): Schedule a call with the Delphina team

## Case Studies

- [Substack](https://delphina.ai/case-studies/substack): "Delphina gives us AI superpowers for data. It's changed how we make decisions." — Chris Best, CEO
- [Basecamp Franchising](https://delphina.ai/case-studies/basecamp-franchising): "I stopped waiting for reports and started finding answers myself." — Zach Gordon, Co-CEO

## Recent blog posts
- [Why Gen AI will transform the workflows of data science and analytics](https://delphina.ai/blog/why-genai-will-transform-workflows) (Dec 13, 2023) — Data science is transformational — it leaves an impact like a crater: profound and enduring. But getting business results from data is still way too hard.
- [The five breaking points for data and AI in the business](https://delphina.ai/blog/breaking-points-for-ml) (Jan 10, 2024) — Deep diving into a question we get all the time from senior leaders: where do data and AI initiatives go wrong?
- [Who should own data and AI?](https://delphina.ai/blog/who-should-own-machine-learning) (Jan 25, 2024) — Today we dive into an uncomfortable question: who owns data and AI?
- [The costliest mistake in data and AI](https://delphina.ai/blog/costliest-mistake-in-machine-learning) (Feb 13, 2024) — Are you solving the right problems? When you don’t get the problem framing right, everything that comes next is a waste.
- [The paradox of data and AI – what leaders need to know](https://delphina.ai/blog/paradox-of-machine-learning) (Feb 28, 2024) — For all the automation it promises, making data and AI work happen is deeply manual. Leaders need a realistic view of what it takes to build data and AI products that deliver value — and how to ensure their teams are actually doing that work.
- [The six most painstaking steps in data work](https://delphina.ai/blog/painstaking-steps-in-machine-learning) (Mar 14, 2024) — If you aren’t involved in the day-to-day work of data and AI, you may assume data scientists spend their time fine-tuning transformer models and performing PhD-level math. Dive in to learn the truth.
- [The seven personas of data and AI](https://delphina.ai/blog/personas-of-machine-learning) (Apr 16, 2024) — Behind the scenes, your team is increasingly worried Data and AI are just a Mirage. Explore the SEVEN key personas on data and AI teams, and the unique challenges they each face in navigating the hype-vs-reality gulf of AI adoption.
- [The danger zone in data science](https://delphina.ai/blog/danger-zone-in-data-science) (May 29, 2024) — Unlike many functions, the returns to quality are highly non-linear in data and AI — and mediocre AI is often downright dangerous. Unpack why, how to identify mediocre AI, and what to do about it.
- [Why PhDs whiff the onsite and how to find a diamond in the rough](https://delphina.ai/blog/interviewing-data-scientists) (Jun 20, 2024) — New PhDs can be total amateurs when it comes to the job market. Knowing these candidates will say some silly things — sometimes unintentionally — how can you separate the wheat from the chaff?
- [What advanced analytics teams are doing that you aren’t](https://delphina.ai/blog/advanced-analytics-with-machine-learning) (Aug 1, 2024) — Data and analytics teams perennially face a burning — yet often unspoken — question: what drives high value actions?

## Recent podcast episodes (High Signal)
- [AI at Planetary Scale: What’s Next for Machine Learning?](https://delphina.ai/podcast/next-evolution-of-ai) (Oct 27, 2023) with Michael Jordan (UC Berkeley)
- [Fooling Yourself Less: The Art of Statistical Thinking in AI](https://delphina.ai/podcast/art-of-statistical-thinking-in-ai) (Nov 1, 2024) with Andrew Gelman (Columbia University)
- [Ramesh Johari on How to Build an Experimentation Machine and Where Most Go Wrong](https://delphina.ai/podcast/ramesh-johari-on-how-to-build-an-experimentation-machine-and-where-most-go-wrong) (Nov 7, 2024) with Ramesh Johari (Stanford University)
- [The Hard Truth About Building AI Systems and What Most Leaders Miss About AI](https://delphina.ai/podcast/gabriel-weintraub-on-the-hard-truth-about-building-ai-systems-and-what-most-leaders-miss-about-ai) (Nov 20, 2024) with Gabriel Weintraub (Stanford Graduate School of Business)
- [Data Science Meets Management: Teamwork, Experimentation, and Decision-Making](https://delphina.ai/podcast/data-science-meets-management) (Dec 1, 2024) with Chiara Farronato (Harvard Business School)
- [What Happens to Data Science in the Age of AI?](https://delphina.ai/podcast/what-happens-to-data-science-in-the-age-of-ai-hilary-mason) (Dec 5, 2024) with Hilary Mason (Hidden Door)
- [What Lies Beyond Machine Learning and AI: Decision Systems and the Future of Data Teams](https://delphina.ai/podcast/chris-wiggins-on-what-lies-beyond-machine-learning-and-ai-decision-systems-and-the-future-of-data-teams) (Dec 19, 2024) with Chris Wiggins (New York Times)
- [From Zero to Scale: Lessons from Airbnb and Beyond](https://delphina.ai/podcast/elena-grewal-on-from-zero-to-scale-building-data-functions-from-airbnb-to-brick-and-mortar) (Jan 9, 2025) with Elena Grewal (Elena's on Orange)
- [Why 90% of Data Science Fails—And How to Fix It](https://delphina.ai/podcast/why-90-of-data-science-fails-and-how-to-fix-it-eric-colson) (Jan 31, 2025) with Eric Colson (Activation Fund)
- [AI Won't Save You But Data Intelligence Will](https://delphina.ai/podcast/ari-kaplan-on-why-ai-wont-save-you-but-data-intelligence-will) (Feb 13, 2025) with Ari Kaplan (Databricks)

## Optional

- [Full content index](https://delphina.ai/llms-full.txt): every page on delphina.ai including all blog posts, podcast episodes, and transcripts
- [High Signal on Spotify](https://open.spotify.com/show/0VewaA4BlDQSyrsmuZShtk)
- [High Signal on Apple Podcasts](https://podcasts.apple.com/us/podcast/high-signal/id1774960199)
- [High Signal on YouTube](https://www.youtube.com/@DelphinaAI)
- [Delphina on LinkedIn](https://www.linkedin.com/company/delphina-ai/)
- [Delphina on Substack](https://delphinaai.substack.com/)
- [Careers](https://jobs.ashbyhq.com/Delphina)