About the roleThis is a hands-on ML role with real end-to-end ownership — you'll take models from idea through to production and own them the whole way. You'll work across some of the highest-leverage systems in the product: the models that help customers price bids accurately, and the systems that match the right subcontractor to the right project (and vice versa). You'll also help build out the agentic systems that are starting to automate the messier parts of the tendering workflow.

• What you'll work on:Take models from idea to production — design, train, evaluate, deploy, iterate — and own them end to end
• Improve the models that help customers evaluate bids and make better procurement decisions, trained on millions of price data points
• Iterate on the ranking, search and retrieval systems that match the right subcontractors to the right tenders — this is a growing focus area for the team
• Design offline evals, run online A/B tests, and drive metrics-led iteration
• Build and improve the agentic system that orchestrates multi-step reasoning across the tendering workflow
• Ship your work into the product — model, backend service, API, and integration into the TypeScript web app
• Maintain production ML and AI systems: monitoring, alerting, continuous improvement

• ✅ Must have requirements:At least 4-5 years building, training and shipping ML models in production — classical ML and deep learning, fluent in a modern framework (ideally PyTorch)
• Experience building ranking, search, recommendation or price prediction systems
• Strong Python — clean, well-structured, production-grade code
• Deep, hands-on experience building and deploying LLM/agent systems: RAG, tool use, multi-step reasoning, evals
• Real experimentation rigor — offline metrics and online A/B tests, letting data decide rather than intuition
• End-to-end product delivery - you've taken things from research code and shipped features, not just handed off models
• Product mindset - you shape what gets built, not just how

• 👍 Bonus points for:German speaker (product conversations are largely in German so this would be very useful day-to-day, but not absolutely required)
• Comfortable shipping your work into a TypeScript/React/Node web app (you don't need to own the front end, but you should be able to contribute to it)
• Embedding models, fine-tuning, or model training at scale
• Domain exposure to construction, procurement or marketplaces
• Experience with GCP or AWS ML infrastructure
• Startup or high-ownership environment in your recent history

About the roleThis is a hands-on ML role with real end-to-end ownership — you'll take models from idea through to production and own them the whole way. You'll work across some of the highest-leverage systems in the product: the models that help customers price bids accurately, and the systems that match the right subcontractor to the right project (and vice versa). You'll also help build out the agentic systems that are starting to automate the messier parts of the tendering workflow.

• What you'll work on:Take models from idea to production — design, train, evaluate, deploy, iterate — and own them end to end
• Improve the models that help customers evaluate bids and make better procurement decisions, trained on millions of price data points
• Iterate on the ranking, search and retrieval systems that match the right subcontractors to the right tenders — this is a growing focus area for the team
• Design offline evals, run online A/B tests, and drive metrics-led iteration
• Build and improve the agentic system that orchestrates multi-step reasoning across the tendering workflow
• Ship your work into the product — model, backend service, API, and integration into the TypeScript web app
• Maintain production ML and AI systems: monitoring, alerting, continuous improvement

• ✅ Must have requirements:At least 4-5 years building, training and shipping ML models in production — classical ML and deep learning, fluent in a modern framework (ideally PyTorch)
• Experience building ranking, search, recommendation or price prediction systems
• Strong Python — clean, well-structured, production-grade code
• Deep, hands-on experience building and deploying LLM/agent systems: RAG, tool use, multi-step reasoning, evals
• Real experimentation rigor — offline metrics and online A/B tests, letting data decide rather than intuition
• End-to-end product delivery - you've taken things from research code and shipped features, not just handed off models
• Product mindset - you shape what gets built, not just how

• 👍 Bonus points for:German speaker (product conversations are largely in German so this would be very useful day-to-day, but not absolutely required)
• Comfortable shipping your work into a TypeScript/React/Node web app (you don't need to own the front end, but you should be able to contribute to it)
• Embedding models, fine-tuning, or model training at scale
• Domain exposure to construction, procurement or marketplaces
• Experience with GCP or AWS ML infrastructure
• Startup or high-ownership environment in your recent history