Harrison.ai, Full Stack Engineer, Generative AI
Jan 2026 – Present · Perth, WA
- Build AI products that help radiologists work more efficiently, cutting the time they spend on reporting and repetitive steps so they can focus on reading scans, working across the React and TypeScript front ends and the Go services behind them.
- Design and maintain an in-house agent orchestration layer that coordinates multi-step LLM workflows, tool calls and state across cloud LLM platforms (including AWS Bedrock).
- Evaluate our automatic speech recognition pipelines against clinical benchmarks, tuning prompt design and post-processing to bring down error rates so radiologists spend less time correcting their dictated reports.
- Prototype orchestration patterns in LangChain alongside our own harness, so we can compare how quickly we can build against how much control we need for clinical work.
- Use Claude Code as a standard part of the daily workflow for scaffolding, refactoring, test generation and code review, cutting time spent on repetitive implementation work.
- Apply core networking and health data protocols (TCP, DNS, HTTP/TLS, MLLP, HL7/FHIR) to integrate agent services with clinical data pipelines carrying sensitive patient information.
- Instrument services with OpenTelemetry for distributed tracing across the agent harness, and work with Kubernetes-based deployments for service orchestration.