SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Tessera Labs is building an AI-powered enterprise transformation platform that helps Fortune 500 companies modernize their legacy systems—SAP, Salesforce, Workday, Oracle, Snowflake, MuleSoft—in weeks rather than years. The company raised $60M (Series A/B) led by Andreessen Horowitz.
As an AI Engineer, you will own the full agent stack that powers this transformation engine. This is not a research or prototyping role; everything you build runs against live enterprise systems that customers depend on for quarter closes.
Key responsibilities:
- Design and ship production agents that understand enterprise landscapes, plan changes, execute across process/data/code, and prove correctness.
- Build the tool layer: typed, permissioned interfaces that let agents act across enterprise systems within human-authorized boundaries.
- Improve agent performance through prompting, context construction, tool-use strategy, and decision logic—diagnosing which lever a given failure requires.
- Build retrieval and indexing over enterprise artifacts: chunking strategies for non-prose content, hybrid search, reranking, grounding, and evaluation.
- Manage context deliberately. Enterprise artifacts are massive; deciding what the model sees, what gets compressed, and what gets dropped is a first-class engineering problem.
- Use classical ML where appropriate (routing, ranking, classification, anomaly detection, confidence estimation) rather than defaulting to LLM calls.
- Build eval and monitoring: task sets from real customer landscapes, regression coverage on every deploy, alerting for schema changes, authorization revocations, or model version issues.
- Instrument every run for auditability: model calls, tool invocations, decisions, and human approvals must be reconstructible.
- Diagnose production failures from execution traces to root cause, then fix the system, not just the prompt.
- Design guardrails, approval gates, and rollback paths that make agents safe in live enterprise environments.
- Build generalizable capability: anything working for only one customer is a product bug, not an engagement feature.
Representative projects include shipping agents that retire legacy custom code, designing approval semantics for live landscapes, building reconciliation agents for merged company data, standing up replay-evaluation pipelines against recorded runs, and turning customer-specific patterns into platform primitives.
You should have 3+ years building and operating production software with meaningful recent time on LLM-powered systems, and ideally have shipped agentic systems that real users depend on.
