AI Technical Lead / AI Architect
First seen by Alion on Sep 26, 2026.
About the Role :
Own the full lifecycle of enterprise-scale AI solutions - architecture through production - and set technical best practices, governance, and standards across the team. A player-coach leadership role.
Key Responsibilities :
- Architect end-to-end DL/RL and agentic AI solutions from design to production.
- Set technical standards, governance, and evaluation frameworks across the team.
- Optimize models for production inference (TensorRT/ONNX/Triton); balance accuracy vs. latency.
- Scale training/inference on Azure ML, AKS/ARO, and distributed infrastructure.
- Lead technical solutioning, manage stakeholders, and mentor engineers.
Must-Have Skills :
- 7+ years ML/AI with production DL and/or RL systems.
- Mastery of DL frameworks (PyTorch/TensorFlow/JAX) and strong applied math (linear algebra, probability, optimization).
- Deep learning across CNNs, transformers, and sequence models; RL agents (PPO, SAC, TD3, CQL).
- Inference optimization (TensorRT/ONNX/Triton) and accuracy vs. latency benchmarking.
- Model serving and containerized deployment at scale (Docker/Kubernetes, AKS/ARO).
- Cloud-scale training/inference on Azure ML with distributed training and MLOps/CI-CD.
- Ability to define standards, governance, and evaluation frameworks across a team.
- Proven technical leadership, mentoring, and stakeholder communication.
- Track record of production deployments with measurable business impact.
Nice to Have :
- Agentic AI architecture (LangGraph, AutoGen, CrewAI) and RAG pipeline design.
- RL libraries (Ray RLlib, Stable-Baselines3, Gymnasium); multi-agent RL and simulation (MuJoCo).
- Model compression/quantization and GPU-efficiency optimization.
- Agentic platform evaluation (Azure AI Foundry, AWS Bedrock, Databricks AgentBricks).
- Optimization/OR background (LP/MIP, Gurobi/CPLEX); Julia/SciML exposure.
- Semiconductor, manufacturing, or supply-chain domain experience.
Skills
AI Architecture, Artificial Intelligence, Machine Learning, Deep Learning, PyTorch, Tensorflow, RAG, Cloud