Software Engineer, Agentic AI

Team Description:

As a Software Engineer on the Agentic AI Team, you'll work directly with the engineers and scientists who use the platform every day. You won't be handed a spec doc from a product manager three layers removed from the problem.

Some days you're heads-down writing harness code, chasing a flaky tool call or shaving minutes off a worker boot. On another day, you might spend a morning pairing with a robotics engineer to figure out why their agent keeps timing out, then spend the afternoon writing a new skill that gives every agent in the company access to a system that didn't exist yesterday. The constant is that you're close to the user and shipping solutions quickly. 

Job Description and Responsibilities:

This role builds the systems that automate digital work at Neuralink. Every day, scientists, engineers, and operators use our agents to perform complex data investigation, query telemetry, debug code, investigate incidents, and stitch data across implants, robots, and clinical trials. 

The platform you'll build is what makes that possible - and what determines how fast the rest of the company can move.

You’ll own a stack that spans dispatchers, persistent worker pools, identity and LLM proxies, session memory, and dozens of skills that connect agents to our internal systems. As more of the company adopts agents, the surface area grows fast - and you'll be the one owning the roadmap and deciding on the highest value projects. You’ll be expected to find new problems at the company to solve.

We're moving fast - model capabilities advance every month, and we're racing to keep our capabilities at the frontier. 60-hour weeks are not uncommon, and there are stretches where the pace is high. This isn't for everyone.

Required Qualifications: 

• A bachelor's degree in Computer Science or equivalent demonstrated work experience
• You've built software that real people use, and you've felt the difference between a demo and a production system
• You can talk to users, understand their workflow, and translate that into a tool an agent can drive reliably. This role is deeply collaborative
• You take ownership: when an agent run breaks in production, you debug it
• You're comfortable with ambiguity:  the patterns for building agentic systems are still being figured out, and a lot of what you build will be the first version of its kind here