Coupa is the platform companies run their spending on – sourcing, procurement, invoices, payments, suppliers, contracts. It is the system of record for how large organisations decide what to buy, from whom, and at what price.
That makes it one of the most unusual datasets in enterprise software: a single network of 10M+ buyers and suppliers, and $10 trillion of transacted spend to date - quotes, bids, awards, orders, invoices, contracts, and the documents behind every one of them. Multimodal, longitudinal, and tied to outcomes measured in real money.
Rossum joined Coupa earlier this year. We brought the document understanding layer - our proprietary T-LLM (transactional LLM) architectures, which we design and train from scratch, and which read the world's messiest business documents in production, millions of them every week. Now we are pointing the same in-house research capability at a much bigger problem: not just reading the documents, but acting on them.
We are expanding our AI Platform team in Prague with a Senior AI Platform Engineer to work on a Sourcing problem set – which suppliers get invited. How the event is structured. How bids that differ in price, lead time, quality, risk and carbon get compared at all. What a fair price even is. When to award, to whom, and how to split the award across suppliers. How to do auctions and autonomous bidding.
This broad problem set will not be solved by a single model. Solutions might come from a diverse set of fields
Recommendation and retrieval
Game theory
Forecasting and should-cost modelling
Combinatorial optimization
Multimodal document understanding etc.
You will collaborate with researchers to implement these heterogeneous workloads in a scalable production platform in the cloud. At the same time, you will design and develop infrastructure for dataset export, model training, and evaluation to allow Research to move fast.
You will work in a small, senior team of engineers - the group that built Rossum's production inference pipeline from scratch - with direct access to Product and the AI Research team. Ideas that work do not sit on paper; they roll into systems used at scale.
What You'll DoLeading the design, implementation, and maintenance of scalable, reliable, and cost-effective AI platform solutions in the cloud.
Collaborate with Research to understand model requirements and translate them into scalable infrastructure.
Improve data pipelines, feature storage, experiment tracking, and model lifecycle workflows.
Build tooling that accelerates experimentation, benchmarking, and reproducibility.
Implement monitoring, observability, and reliability improvements across AI services.
Participate in architectural discussions and contribute to long-term platform strategy.
Partner with Product, Research, and other engineering teams to align platform capabilities with product needs.
Maintain clear documentation and support knowledge-sharing across R&D.
5+ years of experience in product-minded software engineering, ML platform engineering, or infrastructure roles
Proven ability to lead technical teams, maintain team focus, and effectively navigate challenging situations.
Proven track record of delivering ML solutions that drive measurable business and customer impact.
Strong programming experience in a language suitable for ML such as Python
Understanding of distributed systems, microservices, and cloud-native architectures.
Experience with SQL databases (query optimization, database performance tuning, and schema design).
Experience with ML tooling (e.g., experiment tracking, model registries, data pipelines).
Strong problem-solving skills and ability to work in cross-functional R&D environments.
Solid understanding of CI/CD, infrastructure-as-code, and observability tooling.
Internal communication in English as a default.
Experience with training or serving AI/ML models at scale.
Experience with building scalable, automated ETL/ELT pipelines and maintaining robust database architectures (SQL/NoSQL).
Familiarity with data annotation workflows and dataset management.
Experience with GPU workloads, batch/stream processing, or feature stores.
Exposure to Intelligent Document Processing or Deep Neural Network architectures.
We try to keep our stack standardized and minimal: Python, RabbitMQ, S3, Postgres, Triton Inference Server. All deployed with Kustomize and Flux to a Kubernetes cluster in AWS.
We train and deploy our own models: Proprietary T-LLM architectures, designed and trained in-house – not a wrapper around someone else's API.
Transaction volume that brings interesting scaling challenges in the cloud.
Real ownership, short path to customers: You frame the problem, choose the method, and see it working in front of buyers – no research-to-product handoff.
Global impact: Technology used every day by companies around the world.
Experiment-driven culture: Pragmatic delivery, and quarterly recognition for standout research contributions.
Compute and tools: Frontier LLMs on tap for your own work, and our high-end GPU and large-memory clusters to train on.33 days off: PTO, personal days, your birthday and two company wellness days. Parental leave on top.
Prague, Karlín: Inspiring workspace and full tech setup, including a 200 m² terrace with views of Prague Castle.
Skills Required
- 10+ years of experience in applied machine learning, data science, ML engineering, or quantitative research
- Technical leadership experience setting research direction, scoping problems, mentoring others, and owning production outcomes
- Production experience taking machine learning models or decision systems into production
- Strong Python skills
- Experience working with messy, large-scale data
- SQL experience and ability to build trustworthy datasets
- Depth in at least one modeling discipline, such as deep learning, recommendation and ranking, forecasting, optimization, causal inference, reinforcement learning, bandits, market design, or LLM-based systems
- Strong experiment design and scientific rigor, including attention to leakage, baselines, and production-valid evaluation
- Comfort working in ambiguous, greenfield problem spaces
- Interest in procurement, supply chains, or market design
What We Do
Rossum solves four key steps in document-based processes... receiving documents across multiple channels, automated understanding, two-way communication to resolve exceptions, and acting on the data using in-depth integrations. In typical real-world scenarios, Rossum’s proprietary AI engine outranks narrow data extraction solutions in accuracy. Meanwhile, Rossum’s platform automates the document-based communication process end-to-end. Rossum’s goal for every use case is at minimum a 90% document processing speed increase. What does Rossum bring to the table? Zero-friction deployment: See high AI accuracy right out of the box in Rossum’s free trial and cut down on most maintenance effort thanks to cloud hosting and automated self-learning. Highly customizable: Implement powerful configuration APIs while enterprise users can engage Rossum’s dedicated Global Services team. Unified document gateway: Solve everything from security and compliance to IT and user training in one place by adopting a universally capable document solution. End-to-end solution: Rossum’s cloud platform takes care of the entire document lifecycle from receiving to internal IT systems posting. Security and compliance: Rossum is ISO 27001 certified and HIPAA compliant. The cloud service has been specifically engineered for high availability, with enterprise-grade SLAs ranging up to a 99.9% uptime guarantee and 24/7 support
