AI Architect -- Immediate Interview

Job Title: AI Architect 
Location: Austin, TX/Charlotte, NC/San Diego, CA/NYC
Duration: Long Term
Client: Direct
Tech Stack: Python; LLMs on AWS Bedrock (Claude); agentic orchestration (LangGraph or equivalent); code graphs, RAG and knowledge graphs (Neo4j / Amazon Neptune); SQL Server 2024, Red Gate, AWS DMS; Informatica PowerCenter; AWS; CI/CD; and AI-assisted development with coding agents (Claude Code, Cursor).

Job Description: Need to build the Disposition Engine — the reasoning layer that turns deterministic detections (schema drift, ID overlaps, job dependencies, code findings) into high-quality, defensible, auditable decisions at scale, together with the validation harness, the productionization, and the metrics that prove it.

Experience: 
10+ years engineering experience, including significant demonstrated impact designing and deploying production AI systems at scale.
The hands-on technical lead and the one broad role on the pod. You own the Detection & Disposition Engine architecture end to end, you are the senior voice in the joint working sessions, and you write code and review pull requests. Your first job is to convert an open-ended research task into a typed, measurable decision pipeline.

Responsibilities
• Own the end-to-end architecture of the Disposition Engine: the item ledger, deterministic evidence-pack assembly, the rule ladder, the AI reasoning tier, the validation harness, and the closed feedback loop.
• Scale the end-to-end architecture of the existing Detection Engine.
• Decide, per problem, what is deterministic and what genuinely requires reasoning — and defend that boundary. This judgment is the core of the role.
• Design the cross-workstream coupling: how a single schema disposition cascades into batch jobs and code findings, and how coupled items move together as change waves.
• Lead design working sessions; turn ambiguity into a structured decision space and a burn-down the whole team can watch daily.
• Own the gold-set evaluation strategy and the metrics that prove leverage — auto-disposition rate, accuracy against decisions already made by hand, human-minutes per item, and cost per item.
• Drive design-to-code: write production code, review PRs, and oversee testing across the pod.
• Mentor the pod and champion a culture of rigor, velocity and auditable AI.

Qualifications
• Python — production-grade.
• Agentic / multi-agent architecture; orchestration (LangGraph or equivalent), typed state, tool contracts; context and prompt engineering; evaluation engineering.
• Demonstrable judgment on deterministic versus probabilistic system design — you can point to systems where you deliberately kept the model out of the critical path.
• RAG and code/knowledge-graph design; retrieval-pipeline design and tuning.
• Production LLM systems at scale on AWS (Bedrock) or equivalent.
• Evaluation harnesses: golden sets, confidence calibration, regression suites that gate model and prompt changes.
• Spec-driven development; power user of AI coding agents (Claude Code, Cursor, Codex).
• Outstanding written and verbal articulation.