AI Systems Engineer

Thira is building the agentic system of execution for the enterprise: AI agents that autonomously run the back-office work that consumes companies today, across IT, finance, HR, and beyond. We start in Enterprise IT, where repetitive, cross-system, manual work is the norm, and our multi-agent platform resolves the work end to end without human touch. Founded by repeat entrepreneurs with successful exits, including the team that built Apptio, and senior leaders from Meta, Microsoft, Atlassian, IBM, and Oracle.

We are well-funded and moving at frontier‑AI speed.

Thira's agents are only as smart as the systems they learn from. As our AI Systems Engineer, you'll own the knowledge graph that models every enterprise IT environment, and the learning loop that turns each human resolution into a reusable rule, making the platform measurably smarter with every ticket. You'll own the stateful, long‑lived systems underneath the orchestration layer: the part that makes Thira defensible and that competitors can't copy by swapping an API key.

• Design and own the IT knowledge graph, modeling relationships between users, devices, software, permissions, services, and incidents in a queryable, agent‑readable form.
• Build and iterate on the Pattern Miner agent: watch human resolutions, generalize reusable rules, and write them back to the knowledge graph with appropriate confidence scoring.
• Own the Runbook Synthesizer: transform observed resolution patterns into structured, executable runbooks that agents can run autonomously going forward.
• Architect the human review gate and active‑learning pipeline, deciding when a learned rule is trusted enough to run without human approval.
• Design the feedback schema and preference‑data pipeline for eventual fine‑tuning of domain‑specific models on Thira's resolution data.
• Own evaluation for the learning systems: how do you measure whether a rule is correct, and whether a runbook is safe to automate?

• 3+ years building production AI or data systems: knowledge graphs, recommendation systems, learning pipelines, or closely related work.
• Hands‑on experience with graph data modeling: schema design for real‑world domains, not just toy examples.
• Strong understanding of preference learning, feedback loops, or active learning: you've built systems that improve from human signal.
• Familiarity with embedding models, semantic similarity, and retrieval at the knowledge‑graph layer (GraphRAG or equivalent).
• Strong in Python and Java, comfortable with async systems, and rigorous about data quality and schema evolution.
• High bar for correctness in autonomous systems: you think carefully about when a system should trust itself vs. elevate.
• Prior experience building backend systems or full‑stack work.
• Ontology design or semantic‑web technologies (RDF, OWL, SPARQL) in enterprise data contexts.
• Background in ITSM, IT operations, or enterprise infrastructure: you understand what a CMDB is and why current ones fail.
• Prior work on RLHF, DPO, or fine‑tuning pipelines for domain‑specific LLM adaptation.
• Familiarity with graph neural networks or graph‑based retrieval for reasoning over structured enterprise data.
• Experience building evaluation frameworks where correctness is hard to measure (generalization quality, rule precision/recall).

Seattle, WA