Engineering Fellowship

Applications open

Syntony
· Remote
· Durham, NC
· Formation-stage startup

The part-time program offers a defined technical research project for graduate students and people who have completed graduate study. Applications are open by email.

Syntony evaluates frontier and production AI systems and helps clients govern the risks those evaluations uncover. We build software to make testing reproducible and to carry findings through review into control and retesting workflows. Our clients include frontier labs, public-sector organizations, and companies deploying AI.

Syntony is an independent, founder-led organization. Fellows will work with early technical systems and see how research questions become product decisions and scoped engineering work.

The Syntony Engineering Fellowship is a selective, part-time program built around prototype and research-engineering projects in AI evaluation and governance.

Fellows work directly with Nathan Heath, Syntony's Founder & CEO. Each project starts with a defined technical question and bounded scope. It ends with an artifact the fellow can point to when the work is complete.

The format requires independent work between regular working sessions. It is intended for people currently enrolled in graduate school or who have completed graduate study, including researchers and experienced engineers with relevant work to show.

We look for work that shows how you frame a problem, make technical choices, test the result, and explain what you learned.

Projects are scoped so one fellow can own a meaningful piece of work and finish with a usable technical artifact.

You will set the question with Syntony's founder and review technical choices as the project develops. Scope is managed together.

Projects are structured around an artifact the fellow can discuss publicly when confidentiality and intellectual-property terms allow.

Build test harnesses and trace or logging tools that let another evaluator reproduce an adversarial run.

Develop tests for planning, tool use, memory, delegation, and other behavior across multi-step environments.

Build pipelines that collect incident data and score it against a usable taxonomy. The resulting system should surface emerging risk signals.

Prototype ways to turn red-team traces into draft findings and materials for governance review.

Build tools that map feedback loops and escalation paths across a technical system and its organizational or policy environment.

Turn a research question into a working prototype that another person can test and improve.

Eligibility requires current enrollment in a graduate program or completed graduate study. Undergraduate students can apply to the separate Undergraduate Internship .

• Strong programming ability, especially in Python or Type Script.
• Evidence of independent building, such as a Git Hub project, research code, a product or demo, technical writing, or substantial course or research work.
• Clear written communication and the ability to explain technical choices.
• Interest in at least one relevant area, including AI safety, evaluation, governance, red teaming, systems thinking, or strategic risk.
• Sound judgment when requirements are incomplete and when work is confidential.
• Comfort investigating a problem that does not yet have an established solution.

• Experience with LLM evaluation frameworks, agent scaffolds, data pipelines, or developer tools.
• Prior work in AI safety, ML security, policy, forecasting, or…