Want to know if this job is worth applying to?
Seattle, Washington, United States
Hybrid
Not Available
At Sobek AI, we’re building secure AI infrastructure for agentic workflows in life-sciences innovation networks and intergovernmental emergency response. Backed by $10M+ in funding, we work with global, high impact partners on distributed workflows where reliability, security, and trust matter from day one. Our systems are already deployed in mission critical customer environments across 5 countries.
We’re hiring a software engineer to help build the production systems behind Sobek’s core offerings. The work sits where AI powered workflows meet backend services, enterprise data, observability, and product surfaces.
This is a foundational role on the engineering team. We’re looking for someone who has shipped real software used by customers and has hands on experience building AI systems. You should be able to reason about model behavior, data access, evaluation, and production reliability as engineering problems,.
Build AI powered workflows over enterprise and government data, with clear rules for what a model can see, what tools it can call, and when a human needs to review or approve an action.
Design context and grounding systems that give models the right information at the right time without violating permissions or performance constraints.
Work across backend services, APIs, async workers, data pipelines, internal tools, and product facing surfaces.
Design and ship services, data workflows, permissioning, and orchestration components that multiple Sobek products rely on.
Build systems that operate over sensitive data with clear access boundaries, auditability, and predictable performance.
Turn prototypes and customer specific workflows into reusable product infrastructure rather than one off logic.
Build evals and feedback loops for model behavior and workflow outcomes.
Own tracing and runtime visibility across models, context, tool calls, generated outputs.
Debug failures from evidence: context, traces, tool responses, user review, production logs.
Take ownership of important product and platform surfaces without needing heavy direction.
Write clean, maintainable code and create clear abstractions.
Use tools like Claude Code, Codex, and similar systems to move faster, while applying the same standards to generated code as hand written code.
Treat LLMs not as black boxes to call, but as architectural components with failure modes and costs to manage.
You have a track record of shipping production software used by real customers, and you have hands on experience building AI or LLM-powered systems.
You have strong software fundamentals and are fluent in Python and/or TypeScript. Our current stack spans React/TypeScript, Python services, gRPC, AWS, Kubernetes, Terraform, Snowflake, and Docker; exact stack match is less important than range and judgment.
You are comfortable with ambiguity and accompanying ownership.
Especially strong candidates may have:
shipped production LLM or agentic workflows at an AI native startup, scaled AI product company, or serious applied-AI team;
built evals or feedback loops that caught real regressions;
debugged production failures in agent workflows, especially around grounding, tool use, or model/runtime boundaries;
built systems that operate over enterprise data with defined security boundaries;
worked in domains where wrong answers have serious consequences, such as scientific, medical, legal, financial, or public sector workflows;
owned meaningful product or platform surface area earlier than their title would suggest.
Compensation: $170 K – $230 K
Location: Hybrid (Seattle, WA)
Visa: We do not sponsor visas for this role at this time
Benefits: Company-paid health coverage (including dependents)
Equity: Meaningful ownership for early engineers, with flexibility to extend for exceptional scope and impact.
