Head AI Engineer (Full Stack)

Head AI Engineer (Full Stack)

Halcrya Technologies | San Diego, CA | Full-time | In office

Halcrya Technologies is hiring a hands-on technical leader to bring our AI platform for businesses to market. You’ll inherit a well-established infrastructure, work alongside our VP of Engineering, and own AI engineering from the first technical decision through production support.

The mandate is to ship the product to customers within your first 90 days. Our company targets are 10 customers by March 31, 2027, and 100 by December 31, 2027. Your contribution will show up in how quickly customers get value, how reliably the product completes their work, and how much engineering effort it takes to support each additional customer.

The background we’re looking for

• You’ve taken a product through its first commercial launch. You’ve made scope decisions against a deadline, put software in the hands of external customers, and owned what happened next. You can name a feature you cut, explain why the launch could proceed without it, and describe what customers actually used.
• You’ve personally built and operated a production AI system. You can walk through a real request from the user interface to the model, the tools or data it accesses, the action it takes, and the result the customer sees. You can identify the code you wrote, a failure you encountered, and how you verified the fix.
• You’ve inherited someone else’s infrastructure. You know how to read an unfamiliar codebase, trace live behavior, and establish a performance baseline. You can give an example of a system you kept because the evidence justified it and a component you changed because it was holding the product back.
• You’ve led engineers while remaining a practicing engineer. You can break a launch into deliverables, review a difficult pull request, coach an engineer through a design decision, and personally debug a production incident. You leave other engineers able to operate what you build.
• You’ve worked directly with customers. You’ve watched someone try to use your product, discovered where your assumptions were wrong, and changed the software as a result. You can explain the improvement in terms a business owner would recognize.

Technical skills you’ll use

• Full-stack engineering. Build across web interfaces, backend services, APIs, databases, and third-party integrations. Trace failures across those layers and give users a clear way to recover when a workflow stops midway.
• Applied AI and evaluation. Design workflows that use models, retrieval, and tools where they improve the outcome. Build evaluations from real customer tasks, compare performance against a simpler approach, and test changes before they reach users.
• Reliable automation. Handle timeouts, retries, partial failures, and human approvals. If an agent loses its connection after taking an action, you know how to determine what happened and recover without repeating that action.
• Production operations and security. Instrument workflows so the team can explain why a request failed, what it cost, and which customers were affected. Build in access controls, customer data isolation, and audit trails. Work with Technology Operations so releases can be independently verified.
• Technical and commercial judgment. Decide what Halcrya should build and own, what we can buy, and what to remove. Understand how model cost, latency, onboarding time, and support hours affect the economics of serving 100 customers.

What the first 90 days must produce

By day 30, demonstrate the core customer workflow on the existing platform. Agree with the VP of Engineering on launch scope, the biggest risks, and the evidence that will tell us the product is ready.

By day 60, put a working release in initial users’ hands. Observe where they succeed or need help, fix the highest-impact failures, and measure the time it takes a new user to complete the core task.

By day 90, ship the product to market. Customers can onboard and complete the core workflows. The team can measure repeat usage, support the release, and identify what must improve to reach 100 customers.

Questions to expect in the interview

1. Walk us through an AI product you shipped, from a customer’s first click to the final outcome. What did you personally build, and what evidence showed it worked?
2. Which part of that product would a capable competitor have the hardest time reproducing? If both companies had access to the same models tomorrow, what advantage would remain?
3. Tell us about a production failure your tests missed. How did you discover it, how did you protect customers, and what changed afterward?
4. Assume Halcrya has 10 customers and each new account takes a full day of engineering time to onboard. What would you investigate first, and how would you prove you’d removed the bottleneck?

T

o apply, send us one example of an AI product you helped take to market. In a few sentences, tell us what you personally built, the hardest tradeoff you made, and one measurable customer outcome.