Agentic Harness Engineer
DeepJudgeAI-powered Legal company
Zurich, SwitzerlandSenior
Felicis Ventures
Coatue
468 Capital
Andreessen Horowitz
Gokul Rajaram
Robot Ventures
Data & AI
About the role
TL;DR
Develop and maintain the agentic harness for DeepJudge's AI models.
- •Agentic harness is the backbone of our product
- •it transforms powerful models into reliable enterprise-grade applications our customers can trust.
- •You'll join a stellar engineering team and will be responsible for the quality of our agents across applied use cases: developing the harness, crafting system prompts to optimize performance of frontier models, and building evaluation systems that keep a continuous pulse on the reliability and robustness of our features.
- •Key Responsibilities Owning the agentic harness powering critical features across DeepJudge Building and running the evals end-to-end, from design to instrumentation to continuous monitoring Improving agent quality through both code (harness) changes and prompting Influencing our model and launch decisions from the agent quality standpoint Working closely with Legal Engineering, Product, and Customer Success to shape and prioritize harness improvements Requirements Have a Master's degree in Computer Science or a related field, or equivalent practical experience Are proficient in one or more backend programming languages such as Python, Rust, Go, or C++ Know how to evaluate agentic systems and design meaningful eval frameworks Have a strong interest in the latest trends in GenAI and frontier models Like to build things from scratch without hand-holding Have built a large application based on LLMs before
Required skills
PythonLLMsPostgreSQLAirflowPandasNumPyCI/CDKubernetesJestPytestJiraConfluenceFigmaPostmanREST API
Nice-to-have skills
RustGoC++A/B TestingDesign ThinkingWireframingPrototypingUser TestingInformation ArchitectureAccessibility Design
Domain expertise
legaltech
Tech stack
PythonRustGoC++LLMsAirflowPandasNumPyPostgreSQLAWSCI/CDKubernetesGitJestPytestJiraConfluenceFigmaPostmanREST APIgRPCWebSocketsRabbitMQCelerySidekiqNotionSlackAsanaLinearMonday.com
