Senior Software Engineer - Agents

Sage Recruiting Inc.; San Francisco (CA); On-site; USD 150,000 - 300,000; Full time

Job summary

Sage Recruiting's client is building an AI agent for hardware engineering. You will own the core agent layer, enabling long multi-step workflows across locked-down desktop software inside enterprise machines.

This role will execute plain-language requests within tools like SolidWorks, CATIA, NX and Ansys. You'll build tool calling, state management, error recovery, and model routing to keep workflows fast and cheap.

Qualifications

• 3+ years shipping production software.
• Rock-solid CS fundamentals: data structures, algorithms, concurrency and systems design.
• Ability to verify and own AI-written code; not hand-writing code on this team.
• Bias toward shipping measurable results.
• Fluency in any language; Python is a plus.

Responsibilities

• Build tool calling, state management, error recovery, context management and model routing.
• Make it fast and cheap: minimize latency and cost per workflow.
• Build evals from real engineering workflows; define what 'good' means.
• Report directly to the CTO; success measured by task completion and token efficiency.

Skills

Production software

CS fundamentals

Algorithms

Concurrency

Systems design

AI-written code

Python

Tools

COM

Desktop automation

Job description

Sage Recruiting's client is building the AI agent for hardware engineering. Software engineers got Cursor and Claude Code. The people who design cars, planes and rockets are next, and this team is building their version.

Why this company

• The product works today. Their agent runs inside SolidWorks, CATIA, NX and Ansys and does real engineering work from plain-language requests. Customers see about 52% faster engineering and twice as many design iterations.
• The traction is real. Revenue is in the seven figures and has doubled in six months. Fortune 100 automakers, aerospace companies and big tech are paying customers.
• The backing is serious. They've raised a $30M+ Series A from tier-one investors, including early backers of the leading AI labs. Joining before the next round means more equity.
• The bar is high. The founders are technical and have exited companies before. The CTO publishes in Nature and NeurIPS and is a Kaggle Grandmaster, and the tech leads come from Google and Meta.

The role

You'll own the core agent layer, the hardest problem at the company and the reason everything else exists. It takes an engineer's intent and executes long, multi-step workflows across heavyweight desktop software on locked-down enterprise machines.

• Build tool calling, state management, error recovery, context management and model routing
• Make it fast and cheap: latency, cost per workflow, token budgets
• Build evals from real engineering workflows. There's no benchmark for this field yet, so you'll help define what "good" means
• Report directly to the CTO. You're measured on task success, token efficiency and workflow coverage, not clever architecture

No agent experience required. Our client is a software company first. They want great engineers who are curious about agents, not "AI engineers" with shaky fundamentals.

You have

• 3+ years shipping production software
• Rock-solid CS fundamentals: data structures, algorithms, concurrency and systems design
• The judgment to verify and own AI-written code (nobody on this team hand-writes code)
• A bias toward shipping things you can measure
• Fluency in any language. Python is a plus

Bonus points: production agents or eval tooling, desktop automation or COM, a CAD/CAE/PLM background, or open-source and published work.

The details

• $150K to $300K base, plus meaningful equity