🇺🇸Head of US

Stand up the US sovereign instance and run both sides of its market. For a power trader who traded congestion and basis across a continent of nodes and never confused a locational price with a hub.

📍 New York or Houston · Remote-first⬢ The book: US power — nodal LMP across PJM, MISO, CAISO and ERCOT, and the day-ahead / real-time split.◇ Feeds: ISO / RTO real-time LMP

About sovrgn

sovrgn is the sovereign Plan B for AI. Intelligence has become a metered, dispatched, globally-shippable commodity — a second energy market stacked on the first. We run it like one: ask for a capability, not a vendor, and sovrgn resolves the provider under your jurisdiction's law, with automatic failover and fail-closed residency.

It clears on merit order. It splits into firm and interruptible. It has a day-ahead price and a real-time one, and it prices by where the work lands. If that sentence reads like home, keep going — you already speak this market.

Australia's instance is live; the AloomU cluster is its sovereign-silicon backstop. The other national instances open next, and each one needs someone to run it.

The book you ran

US power across one or more ISOs/RTOs: day-ahead and real-time energy, FTRs/CRRs, congestion and basis across thousands of nodes. You know a locational marginal price is not a hub price, and you've been on the wrong side of a constraint binding.

Locational pricing for inference is the same instinct — where the work clears is where it's cheapest, minus the transmission loss electrons pay and tokens don't. Merit order, firm vs interruptible, day-ahead vs real-time: you already trade all three.

What you'll do

  • Stand up the US sovereign instance end to end — providers, residency boundary, and the on-ramp.
  • Own the buy side: bring US enterprises and public-sector buyers onto the instance and grow demand.
  • Own the sell side: bring in-jurisdiction capacity onto the market — sovereign silicon, spare GPU, firm and floating.
  • Run the market itself: firm vs floating pricing, failover chains, and the residency rules that keep US data in-country.
  • Be the face of sovrgn in the US — enterprise buyers, insurers, and regulators across a fragmented, state-by-state landscape.

What we're looking for

  • You have personally carried a US power or gas book — a trading desk, a utility/IPP, a co-op, a retailer, or a large industrial/data-center load.
  • Fluent in at least one ISO/RTO (PJM, MISO, CAISO, ERCOT, SPP, ISO-NE, NYISO): day-ahead, real-time, congestion, and FTRs/CRRs.
  • You settle debates with numbers. Nodal LMP, basis, heat rates, and the cost of certainty are second nature.
  • Commercial range: you can talk to a CIO, a developer, and a state regulator in the same week.
  • AI-native or AI-curious. You don't need to write the router — you need to understand the market it runs.
  • US-based. Comfortable that the market is fragmented and the law varies by state.

Bonus points

  • Origination/structuring — PPAs, tolling, data-center power deals.
  • ERCOT or data-center-load experience (where AI demand and power demand are already the same conversation).
  • You've built a desk or a book from scratch, not just inherited one.

Why this desk

  • First in seat for your market. You're not inheriting a playbook — you're writing it, the way you always wished your old desk had let you.
  • A market you already understand, against a new commodity. The mechanics are the ones you traded: merit order, firm vs interruptible, day-ahead/real-time, locational pricing. We just changed what's being dispatched.
  • Direct line to the founders. Decisions get made in hours, not weeks.
  • Equity for the right person.

How to apply

Use the form at the bottom of this page — tell us which book you ran and where. A short note on the trade you're proudest of (or the one that taught you the most) goes a long way.

We read every application but won't reply to everyone — if we want to talk, we'll reach out.

🇺🇸 Apply · Head of US

Apply for Head of US.

Tell us where you traded and which book you ran — we'll reply directly.