Modern Relay is building the knowledge platform for the agent era. Our product caters a new kind of company: one in which humans work alongside internal and external AI agents, and where coordination, context and trust become critical infrastructure. The platform provides a shared layer of truth where both humans and agents can propose updates, contribute knowledge and trigger workflows. This result in a living, compounding knowledge hub that can be read from, written to and improved by both people and software.
Role OverviewWe’re looking for an AI Engineer to help build the data and model foundations that make Modern Relay’s platform reliable in production. You’ll work across data pipelines, model development, and ML infrastructure, turning messy signals into structured knowledge and high-quality model behavior. This role is ideal for someone who enjoys shipping end-to-end systems, from schema design and data infrastructure to training/evaluating models and improving them with feedback loops.
LocationsSan Francisco, CA
New York City, NY
Barcelona, Spain
Remote (U.S. and Europe)
Design and build data pipelines that ingest, clean, and transform product and customer data into high-signal training and evaluation datasets
Own data infrastructure decisions (storage, orchestration, lineage, observability) to ensure reliability, scalability, and fast iteration
Develop and improve ML/AI systems that power agent's behavior in task-solving, including retrieval, ranking, classification, and structured extraction
Create and maintain schemas for agent memory, tool outputs, and conversation artifacts to make downstream modeling and analytics consistent
Build evaluation harnesses and metrics to measure model quality, regressions, and real-world performance (offline + online)
Work with knowledge representations (e.g., knowledge graphs) to connect entities, events, and business context for better reasoning and retrieval
Partner closely with Product and Engineering to integrate models into production workflows with clear SLAs and monitoring
Continuously improve feedback loops: labeling strategies, active learning, error analysis, and dataset/version management
Data pipelines and datasets are trustworthy, well-instrumented, and easy to iterate on as product needs evolve
Model performance improves measurably over time with clear evaluation methodology and fast debugging cycles
Agent outputs become more consistent and structured through strong schema design and robust post-processing/validation
Knowledge and retrieval systems reduce hallucinations and increase task completion rates in real customer workflows
Cross-functional teams can confidently ship AI improvements because quality, monitoring, and rollback paths are in place
0–6 years of experience in AI/ML engineering, data engineering, or a closely related role (we’re open to exceptional new grads with strong projects)
Strong fundamentals in data engineering: pipelines, data modeling, schema design, and data quality practices
Experience building or operating ML systems in production (training, evaluation, deployment, monitoring) or strong evidence you can ramp quickly
Comfort working across the stack: from raw data and infrastructure to model behavior and product integration
Familiarity with modern ML platforms and tooling (experiment tracking, dataset/versioning, orchestration, feature/data stores, model serving)
Understanding of information theory concepts (e.g., entropy, mutual information) and how they relate to signal, compression, and evaluation
Experience with knowledge graphs or structured knowledge representations is a plus
High ownership and a bias toward shipping: you can take ambiguous problems, propose a plan, and execute
Data pipelines
Data engineering and data infrastructure
AI / artificial intelligence
Machine learning platforms and production ML
Model development, evaluation, and monitoring
Schema design and structured data systems
Knowledge graphs and information retrieval
Information theory fundamentals
Build core AI infrastructure that directly impacts product reliability and customer outcomes
Work on real-world agent coordination problems where data quality, structure, and evaluation matter as much as models
High autonomy and ownership in a fast-moving team shipping at the frontier of applied AI
A chance to define how Modern Relay’s agents learn from data and improve over time
