One founder. One thesis. The data platform for the decade after the data warehouse.
A solo-founder bet on what the next $100B data company looks like — built before the AI-agent era forces every incumbent to rewrite themselves.
The honest version.
I've spent a decade inside the data platforms that this one is built to replace. I ran AI R&D inside one of the largest Snowflake consulting partners. I have seen, up close, what it costs a Fortune 500 to live inside a consumption-priced warehouse with proprietary tables, retrofitted governance, and agent capabilities bolted on after the fact. Customers do not love any of it. They tolerate it because the migration cost is real and the alternatives are worse.
The alternatives were worse. They are not anymore. Apache Iceberg is now a real open format. MCP is now a real agent protocol. GPU-native query execution is finally production-real on the right workloads. Multi-engine routing has crossed from theoretical into shippable. Compounding intelligence with privacy-preserving aggregation is no longer a research paper.
Snowflake was the right architecture for 2014. Databricks was the right architecture for 2018. Neither was designed for the world where agents are the primary users of data, where compliance is enforced code rather than PDFs, where the customer wants the substrate portable, and where intelligence compounds across customers under a privacy-preserving floor. They cannot become Eisberg without rewriting themselves — and they will not, because it would cannibalize the consumption model that funds the companies.
So I'm building it. Solo, for now, with the platform far enough along that it shipped 1,200+ tests green and 27 sessions of substantive engineering before the marketing site went live. Design partners coming online over the next two quarters. SOC 2 Type II in progress. Customers — when they sign — will be people who feel the pain in the bill and the audit log every quarter.
That is the bet. If you are an engineer who has felt this pain, a CISO tired of getting promises about agent governance, or a CFO doing the math on next quarter's Snowflake bill — I'd like to talk.
Six commitments that will not break.
These exist so a buyer, a regulator, or a future hire can hold the company to specific behavior — not vibes. If any of these break, it is fair to call it out publicly.
Honesty over polish.
I will never claim a capability we cannot demonstrate end-to-end on your data. I will never bill for a speedup we did not measure. Marketing-grade numbers are how trust dies. If a benchmark hasn't been run on customer hardware, the slide doesn't say it.
Customer-owned data, structurally.
Eisberg is built so we are incapable of holding your data. The day you decide we are not the right partner, your data is already where it needs to be — in your bucket, in open Iceberg, behind your KMS keys. Most incumbents cannot match this without abandoning their revenue model. We started here on purpose.
Open formats are the only formats.
Apache Iceberg for storage. MCP for agents. Open-standard catalog and lineage protocols. Vendor-format gravity is the lock-in mechanic of the previous era. We are not perpetuating it, and we will not acquire our way into it.
Agents are the primary user.
By 2030, agents will issue more queries against enterprise data than humans. Eisberg was designed for that world from line one — MCP-native, identity-bound, signed Birth Certificates, per-action metering. Not bolted onto a human-first warehouse.
Compounding intelligence is the moat.
Every classification, every query optimization, every tribal rule captured makes the platform smarter. Across customers (with consent, k≥3 anonymity gate) and across time. The longer it runs, the wider the gap. The 500th customer benefits from the lessons learned from the first 499.
Eisberg will not be acquired by a hyperscaler.
The whole architecture is multi-cloud and customer-owned. Selling that to AWS, Azure, or GCP would mean breaking the promise the platform is built around. The plan is a public-company data platform, or nothing.
Want to talk?
Demo, design partnership, FinServ pilot, security review, technical question — I respond to every email within one business day. No SDR funnel, no marketing automation, no calendar bot. Just a human reading your message.