Principal Engineer (AI/ML Architecture)
Posted Updated
Description
- Coupa’s AI platform powers production agents, conversational AI (Navi), customer-built agents (Agent Studio), and a suite of ML models for classification, fraud detection, and supplier intelligence
- The Principal Engineer, AI/ML Architecture will define the technical direction for the next generation of these capabilities, designing how we train, evaluate, and serve models that are deeply tuned to enterprise spend management
- Reporting to the Sr. Director, you will be hands-on in architecture and prototyping while guiding a growing team of ML and data engineers
- Define the architecture for model training, evaluation, and serving across Coupa’s AI platform
- Evaluate and select model approaches (open-weight, commercial, and hybrid) against enterprise accuracy and cost requirements
- Design evaluation frameworks that measure model quality on Coupa’s specific task categories
- Drive technical partnership evaluations with AI infrastructure and model providers
- Architect training data pipelines, including synthetic data generation and quality validation
- Design retrieval-augmented generation (RAG) systems that extend our existing RAG infrastructure with structured knowledge retrieval
- Establish technical standards for model safety, tenant data isolation, and responsible AI
- Write code, review PRs, and prototype approaches, especially in the early phases
- Mentor and guide ML and data engineers across US and India
- Collaborate with product, existing AI platform, and cloud operations teams
- Experience with enterprise B2B SaaS platforms preferred
- Experience designing and running custom evaluation suites for LLMs
- Hands-on experience with cloud AI/ML services (model hosting, managed training, or equivalent)
- 15+ years of software engineering experience, with 5+ years focused on ML/AI systems
- Advanced degree in Computer Science, Machine Learning, or equivalent practical experience
- Deep knowledge of transformer architectures, training optimization (LoRA, QLoRA, PEFT, RLHF, DPO), and inference serving
- Demonstrated experience training or fine-tuning large language models. Must have shipped a fine-tuned or domain-adapted model to production
- Strong understanding of RAG architectures, vector search, embedding models, and knowledge graph integration
- Experience with distributed training on GPU clusters
- Proficiency in Python, PyTorch, and ML infrastructure tooling