About Kanan
Kanan is building AI software to help manufacturers and financial institutions understand supply chains, assess risk, and make informed decisions. We are bringing together a founding team across AI research, engineering, and data science.
About the role
As a Founding AI Engineer focused on applied AI, you will build the intelligence components behind Kanan’s software. You will turn research ideas into working systems that process documents, retrieve relevant information, connect facts, and support business analysis.
You will work closely with the CSO on scientific approaches and with the Full Stack Engineer on deployment and integration. This role combines strong coding with disciplined experimentation: implementing a method, measuring how it performs, understanding where it fails, and improving it. Your work will help make AI outputs useful and traceable when source information is fragmented, incomplete, or contradictory. You will maintain clear records of experiments and reusable evaluation tools so improvements can be checked and reproduced by the rest of the team.
Responsibilities
- Develop AI agents, document-processing pipelines, and retrieval systems that work with a mix of structured records and unstructured business information.
- Build knowledge-graph and information-extraction components that connect entities and relationships while preserving links to the sources behind them.
- Implement and compare models, research methods, and reasoning approaches, using reproducible experiments to understand performance and practical tradeoffs.
- Create evaluation datasets and tests for accuracy, grounding, and failure handling, including difficult cases where the system should request review.
- Work with the CSO and Full Stack Engineer to deploy validated capabilities, investigate errors, and improve quality, latency, and cost over time.
Required qualifications
- Strong Python and software-engineering skills, with experience building maintainable components, writing tests, and debugging systems that use multiple data sources.
- Hands-on experience with language models, retrieval, and agent workflows, including examples that go beyond a demonstration or isolated notebook.
- Understanding of machine learning, experimentation, and model evaluation, with the ability to choose useful metrics and interpret results critically.
- Ability to implement research ideas, explain their strengths and limitations, and collaborate with researchers and product engineers to deliver practical results.
Preferred experience
- Experience with entity resolution, knowledge graphs, document AI, or extracting structured information from complex documents.
- Experience with model adaptation or combining machine learning with structured rules, especially where outputs need to be reviewed and explained.