Senior AI Engineer

Colt Technology Services

96d•Hybrid

About The Position

Colt Technology Services is a global digital infrastructure company, operating one of the world's largest fiber networks across Europe and Asia. We are building a Boulder-based AI team as the engineering center of our AI Practice. The team is responsible for how AI is adopted, governed, and scaled across Colt. The AI Practice operates across several parallel workstreams, including use case delivery, AI WAN, and private AI. The Boulder team is where we build, test, and validate before we scale. This is a build role. You will take AI use cases from a business problem to something running in production and being used every day: retrieval and search over Colt's own data, agents that carry out multi-step work, and AI built into the systems Colt's teams already rely on. We are looking for a strong software engineer first. If you have spent eight to ten years building and running production systems and even if you have not yet shipped an LLM-based product, we want to hear from. The platform, the tooling, and the model access are already in place, and a staff-level AI engineer on the team will pair with you directly on the AI side of the work. The AI-specific skills you can bring to the table is a plus but is also learnable. The engineering experience and judgement that comes from having owned real systems in production is not, and that is the part we are hiring for. What we do expect is that you have already got hands on with AI, whether at work or as a side project for yourself. You will be working directly with the Colt teams taking requirements from them. The team is small by design, the problems are real, and what you build in your first year sets the pattern for how Colt builds AI.

Requirements

8-10+ years building production software. You have owned systems end to end: designed them, shipped them, operated them, and been on the hook when they broke.
Strong in Python, or strong in another language such as Java, Go, C#, or TypeScript and ready to work primarily in Python.
APIs and service integration, working with real data at scale, testing, CI/CD, and debugging things that are already live in front of users.
Comfortable in any major cloud. We build primarily on Google Cloud; equivalent AWS or Azure experience is ok to start.
Comfortable with ambiguity and with direct contact with non-technical stakeholders. You ask what the problem is before deciding what to build.
You have already built something with it under your own steam. A side project, a prototype at work, an internal tool, a serious evaluation you ran, or agentic coding tools you use daily and have opinions about. It does not need to have shipped, been finished, or worked. We want to hear what you tried, what broke, and what you concluded.

Nice To Haves

Any of this at enterprise scale: production RAG over messy real-world data, agent frameworks, evaluation tooling, Vertex AI, Kubernetes, Terraform. Useful if you have it, and teachable if you do not.

Responsibilities

Design and build AI applications end to end: data access, retrieval, model calls, orchestration, and the interfaces people use
Work with large language models through APIs and self-hosted inference, covering prompting, tool use, retrieval, and multi-step agent workflows
Build evaluation harnesses that show whether a use case is good enough to ship, and keep owning them after it ships
Integrate AI into Colt's existing systems and data, within the constraints of a live enterprise environment
Work with the Colt teams, who will be your customers, and who will use what you build, across network operations, service delivery, finance, and others
Turn a loosely described business problem into a scoped, buildable design
Show working software early and often, and change direction when the first answer turns out to be wrong
Be straight with stakeholders about what AI is not the right tool for
Move your own work from prototype to production-grade: tested, instrumented, documented, and supportable
Monitor accuracy, latency, and cost in production, and act when they drift
Work with platform engineers on deployment, and hand over cleanly to production operations
Contribute reusable patterns, components, and internal libraries so the next use case is faster than the last