Senior Internal Automation Engineer

ABOUT VOLTA

Volta is the category-defining, fully vertically integrated AI infrastructure platform – from capital to clusters to software, under a founder-led enterprise. Our mission is The Utility of Compute™: AI infrastructure as dependable and available as electricity, for every organization that needs it. Launched with a $10B strategic partnership with one of the leading frontier AI labs, a Series A led by Andreessen Horowitz, and a $5B AI Infrastructure Fund, Volta is building the infrastructure layer of the AI era from the ground up. We are 100+ people across London, Palo Alto, and New York, with rapid growth expectations to hundreds.

ABOUT THE ROLE

Volta builds and operates large scale GPU compute infrastructure for frontier AI customers. Growing this fast produces a long tail of manual internal work: invoices retyped into banking portals, expense reports chased by hand, spreadsheets reconciled every month, accounts provisioned ticket by ticket. Each task is small. Together they cost the company real time, and they multiply with every new sit and every new hire.

This role removes that work permanently. You sit with the person who owns a process, understand what it actually does, then build the integration that runs it against banking and accounting systems, Microsoft Graph / Office, Slack, Jira and Confluence, and the HR and identity stack. Most of the answer is well built API code with sound authentication, error handling, escalation, audibility and observability. Some of it uses models, for classification, extraction, or matching, where that is the right tool rather than the fashionable one. Knowing the difference is a large part of the job.

Much of this work lands inside the financial reporting path, and Volta is building toward SOX compliance alongside its existing ISO 27001 and SOC 2 obligations. An automation that touches invoices, payments, or reconciliation becomes part of the control environment the moment it goes live. That sets the bar: changes are reviewed, tested, and traceable, access is least privilege and segregated, every run leaves an audit trail an external auditor can follow, and nobody moves code into production by hand. You build to that standard from the first commit rather than retrofitting it under deadline.

Regulation shapes the work in a second way. Volta is headquartered in the UK and operates across the EU, so anything you build with AI in it sits under the EU AI Act as well as UK and EU data protection law. Automation that touches hiring, performance, or other employment decisions carries the heaviest obligations, and the line between a helpful classifier and a regulated decision system is easy to cross without noticing. You are expected to recognize where that line runs, keep a person in the loop where the law requires one, document what you deploy, and bring Legal and Security in early rather than after the build.

You sit inside Corporate IT at a fast-growing AI-forward company, which owns the tooling, accounts, and API access you build on, and which gives you engineers next to you for design discussion and code review rather than leaving you to mark your own homework. Security Engineering advises on access and compliance. The first internal customer is the accounting and finance team. After that the scope is the whole company: finance, IT, People, delivery, product, and the engineering teams for their own internal overhead. Very little here is customer facing. Initially, you are the only person doing this work full time, so you prioritize hard, ship small, and build things that keep running without you next to them.

WHAT YOU WILL BE DOING

• Automate finance operations first: ingest received invoices into banking and bookkeeping systems, classify and match them, and close the loop on the exceptions

• Work directly with process owners across finance, IT, People, and delivery: take a described workflow and turn it into a specified, testable automation

• Work as part of Corporate IT: review their changes, have yours reviewed, and keep automation aligned with how the underlying systems are administered

• Build and run integrations against internal systems, including Microsoft Graph, Slack, Atlassian, banking platforms, and HR and identity tooling

• Own authentication and credentials for those integrations: OAuth flows, service principals, scoped tokens, secret rotation

• Build every finance touching automation to withstand audit: version controlled and peer reviewed changes, automated deployment with no manual production edits, least privilege service identities, segregation of duties preserved rather than automated away, and retained evidence of what ran, when, and on whose authority

• Work with Finance and Security Engineering on control design, so an automation replaces a manual control with a stronger automated one instead of quietly removing it

• Apply models where they earn their place, for document extraction, classification, or matching, and use conventional code everywhere else. Keep a human decision point where a control requires one

• Build agentic automations where a single scripted path will not hold: an agent that picks up a case, calls the tools it needs across Microsoft Graph, Slack, Atlassian and the finance stack, and stops at a human approval gate before anything with money or access impact. You define the tool surface, the permitted scope, and the blast radius, rather than handing a model a broad credential and watching what happens.

• Keep AI use inside its regulatory boundaries: automatically classify use cases correctly under the

• EU AI Act, apply transparency and human oversight where required, document deployed systems, and escalate anything touching employment decisions to Legal before it is built

• Instrument what you build so failures surface immediately and the state of any run is visible without reading logs

• Maintain what you deploy, including the unglamorous part where an upstream API changes and the process still has to run on Monday

• Set the patterns, repositories, and review practices that keep internal automation maintainable and auditable as more of it accumulates

• Push back when a process should be simplified or deleted instead of automated

WHAT YOU BRING

• Production software development held to the standard of an engineering team, not scripting attached to an operational role

• Fluency across more than one software ecosystem. Internal systems arrive in whatever language, runtime, and SDK their vendor picked, and you are expected to work in what the problem needs rather than bending every problem toward one stack. Python and Go come up often here, neither is a requirement

• Deep practical work with third party APIs: pagination, rate limits, idempotency, retries, webhooks, and how an integration behaves when the far end is having a bad day

• Real command of authentication mechanisms, including OAuth 2.0, OIDC, service accounts, and API key and secret management

• Disciplined engineering practice that produces an audit trail as a by-product: source control, code review, automated testing, controlled deployment, and clear separation between who builds a change and who releases it

• Hands on experience building with LLMs, including tool use and structured extraction, plus the judgment to recognize when a model is the wrong answer

• Hands on work with agentic patterns: tool and function calling, MCP or equivalent tool interfaces, state and retry handling across multi step runs, and testing agent behavior against realistic cases before it touches a production system.

• Working awareness of the rules governing AI and automated decision making in a workplace context, at the level of knowing which use cases need legal review before a line of code is written

• Ability to sit with a subject matter expert, understand a business process well enough to implement it correctly, and translate it without needing them to specify the solution

• A track recor