This list contains only the countries for which job offers have been published in the selected language (e.g., in the French version, only job offers written in French are displayed, and in the English version, only those in English).
At Valtech, you’ll find an environment designed for continuous learning, meaningful impact, and professional growth. Whether you're pioneering new digital solutions, challenging conventional thinking or building the next generation of customer experiences, your work will help transform industries.
Job Responsibility
Architecture & Systems Design: Design and own event-driven, micro-services architectures on GCP
Architect end-to-end RAG pipelines - ingestion, chunking, retrieval, and generation - for production workloads
Define and enforce network and access security patterns: IAP, VPC-SC, Secret Manager
Author Architecture Decision Records (ADRs) and enforce Definition of Done (DoD) standards
Trade-offs, Cost & Performance: Lead LLM selection decisions - Vertex AI vs. open-source - with structured cost-benefit framing
Optimize GenAI run costs through token budgeting, embedding cache strategies, and model tiering
Own scalability design for Cloud Run workloads, including cold-start mitigation and load-testing strategy
Translate architectural trade-offs into clear stakeholder communication
Data & Document Quality: Design source modeling strategies across Drive, GCS, and Vector Search
Define and enforce chunking strategies and Registry governance for document pipelines
Proactively identify and retire technical debt
maintain a prioritized remediation backlog
Technical Leadership: Mentor engineers through code reviews, pair programming, and design walkthroughs
Serve as the primary technical interface with FR Core Team and AI sponsors
Champion engineering standards and drive adoption of best practices across the squad
Requirements
8+ years of software engineering experience, with 3+ years in AI/ML engineering or GenAI platform roles
Production-grade RAG system delivery - from ingestion through retrieval to generation - not just prototyping