Infrastructure Engineer, Applied AI

P2P; San Francisco (CA); On-site; USD 210,000 - 290,000; Full time

Job summary

Paradigm, a San Francisco-based investment firm focused on frontier technologies, is hiring an Infrastructure Engineer for Applied AI. You will build and maintain core services that coordinate agent runs, ensure durable state, and harden a Kubernetes-based sandbox runtime for secure AI workloads.

You will also manage secure credentialing for API access, maintain reliability and observability, and enable self-hosted deployment of Centaur across organizations.

Qualifications

• A track record of building world-class products.
• Experience applying modern AI and software systems to solve complex, ambiguous problems end-to-end.
• Comfortable navigating everything from foundational models to full-stack product delivery.
• Pragmatic, product-minded builder who cares about solving real problems rather than overfitting to academic ML goals.
• Ability to work autonomously and as a collaborator on the team.

Responsibilities

• Control plane and durability: Own the core services that coordinate agent runs, keeping execution state durable and recoverable through restarts, failovers, and everything that goes wrong in production.
• Execution runtime: Build and harden the sandboxed, Kubernetes-based runtime that isolates agent sessions, covering provisioning, lifecycle, and the network controls that make untrusted agent code safe to run.
• Secrets and secure egress: Maintain the credential model that lets agents call third-party APIs without ever holding raw keys.
• Reliability and observability: Own monitoring, alerting, incident response, and the failure-recovery behavior the system relies on.
• Self-hosted deployment: Keep Centaur something any organization can run inside its own walls, with clean deployment, sane configuration, and painless upgrades.

Skills

Product development

AI systems

Foundational models

Pragmatic product mindset

Autonomous / collaborative

Tools

Kubernetes

Job description

INFRASTRUCTURE ENGINEER, APPLIED AI

The Firm

Paradigm is a San Francisco-based investment firm focused on frontier technologies across the globe, with over $11 billion in assets under management. We make investments in companies and protocols at all stages, ranging from early-stage venture financing rounds to growth equity to liquid assets.

Paradigm was co-founded in 2018 and is co-led by Matt Huang and Alana Palmedo. We've been hard at work investigating the world's most beautiful technical problems since day one. Our research-driven approach helps us build relationships with founders and entrepreneurs, but it also reflects our broader goal of accelerating the progress of frontier technology.

The Role

Over the last eight years Paradigm has built some of the most used open-source software in the crypto industry, including Reth and Foundry. Earlier this year we published EVMBench with OpenAI, launched Optimization Arena and hosted the Auto Research hackathon. More recently, we open-sourced Centaur, a self-hosted runtime for multiplayer AI agents, which we've been using to transform how we work at Paradigm and Tempo.

All of these were done part-time by people on Paradigm's Investing & Research team. This has made us realize that we are people-bound, not idea or impact bound. We need to grow our team to achieve our ambitions.

We want to do three things:

1. Experiment aggressively. We will push the capabilities of models, harnesses and infrastructure to the limit in service of Paradigm's broader goals in investing, research, and building.
2. Build in public. We will get our hands dirty building useful open-source products.
3. Invest. We will leverage our infrastructure - an area we think is still ripe for disruption with AI - to be better investors. By understanding infrastructure gaps experientially, we may incubate companies that will accrue value by addressing these gaps.

As we deploy more long-running agentic capabilities, we may get into fundamental research, forecasting, model cost optimizations etc., but that is not our focus yet.

We are looking for fearless engineers for this project who understand infrastructure, security, AI models, harnesses, and Slackbots. You will work directly with Georgios, Arjun, Matt Slipper, and the broader Paradigm team, alongside Tempo, and other companies that are themselves operating at the frontier.

Your work will directly impact how Paradigm operates every day.

Responsibilities

• Control plane and durability: Own the core services that coordinate agent runs, keeping execution state durable and recoverable through restarts, failovers, and everything that goes wrong in production.
• Execution runtime: Build and harden the sandboxed, Kubernetes-based runtime that isolates agent sessions, covering provisioning, lifecycle, and the network controls that make untrusted agent code safe to run.
• Secrets and secure egress: Maintain the credential model that lets agents call third-party APIs without ever holding raw keys.
• Reliability and observability: Own monitoring, alerting, incident response, and the failure-recovery behavior the system relies on.
• Self-hosted deployment: Keep Centaur something any organization can run inside its own walls, with clean deployment, sane configuration, and painless upgrades.

Qualifications and Experience

• A track record of building world-class products
• Experience applying modern AI and software systems to solve complex, ambiguous problems end-to-end
• Comfortable navigating everything from foundational models to full-stack product delivery
• Pragmatic, product-minded builder who cares about solving real problems rather than overfitting to academic ML goals
• Ability to work autonomously and as a collaborator on the team

Attributes

• Exceptional team player
• Extreme open-mindedness
• Clarity of thought
• Clear and concise communication (both written and verbal)
• Technical depth; analytical; rigorous
• Highly curious; fast learner
• Ability to bridge technical and investing mindsets
• Interest in frontier technologies and crypto markets