We are looking for a working student to join the Janus team and help build the data and AI infrastructure powering our freight intelligence platform. This is a hands-on role with real ownership from day one.
In the first months, you'll focus on benchmarking and evaluating AI models against real freight data. This means testing local and cloud models — including DeepSeek, DeepSeek OCR, GPT-4, and Claude — across a range of freight documents: tariffs, surcharges, waybills and bills of lading, invoices, packing lists, purchase orders, and spot market rates. You'll build dashboards to track and compare model performance and help us determine the right model architecture for each use case.
On the development side, you'll build web scrapers to extract spot market rates and carrier data — including sailing schedules and service information — from carrier websites. You'll also contribute to a knowledge library that feeds our AI agent infrastructure.
A smaller portion of your time will go toward internal tooling, supporting agent optimization and MCP pipelines.
You are pursuing or have completed a degree in Computer Science, AI, or a related field. You write clean Python and are comfortable with Typescript for web scraping. You have hands-on experience with AI models beyond prompting — ideally with DeepSeek — and understand what it means to deploy, benchmark, and evaluate model performance in a real environment.
We are a small, fast-moving team building financial and operational infrastructure for the global freight industry. You will work directly with the founders, ship real features, and see your work used in production quickly.