AI Architect
First seen by Alion on Sep 10, 2026.
Job Description :
- Own the architecture, and make it easy to build on.
- Own and evolve our enterprise reference architecture for generative and agentic AI - across models, retrieval, orchestration, reasoning, and the governance that runs through all of it.
- Build and demonstrate, hands-on. Prototype in code, stand up working proofs and demos, and stay close to the models and tools - leading by showing, and learning continuously as the field moves.
- Build reusable frameworks, blueprints, and paved-road patterns that engineering and product teams adopt.
- Lead solutioning from proof-of-concept to production: scope experiments, prove them, and build the roadmap to scale.
- Stand up the shared AI platform and enablement - model serving, vector and graph retrieval, tool and agent orchestration, observability, and evaluation - and help teams build on it safely.
- Embed governance by design: traceability, human-in-the-loop, evaluation, and tenant isolation built into every architecture from the start.
- Own build-vs-buy decisions and set technical standards; run architecture reviews and mentor engineers.
- Translate technical trade-offs clearly for leadership, clients, and engineering.
WHAT YOU'LL BRING :
- Production-proven, end to end.
- Deeply hands-on - you still build. 8+ years in software / AI engineering, with production-proven delivery of agentic, generative, and classical ML/DL systems - not just POCs.
- Hands-on depth across the modern agentic stack: orchestration frameworks (e.g. LangGraph, Dapr Agents), protocols (MCP; familiarity with A2A), RAG and GraphRAG, vector and graph databases (e.g. Neo4j, pgvector), model serving (e.g. vLLM), and LLMOps (versioning, drift, cost, evaluation).
- Practical evaluation and guardrails experience - making model quality measurable and defensible.
- Command of the full span from classical ML/DL to generative to agentic, and the judgment to choose the simplest approach that works.
- Cloud and data-platform depth (AWS / Azure / Databricks) and familiarity with semantic layers and lakehouse architectures.
- End-to-end architecture ownership and the ability to set and evolve standards across an organisation.
- Excellent communication and stakeholder management, including with senior leadership, and a track record of mentoring.
Skills
Artificial Intelligence, Machine Learning, Data Science, Data Scientist, LangChain, Python, MLOps