Job Description :

You'll make an impact by :

- Agent harness architecture - design and own the runtime that wraps LLMs: orchestration, tool integration, memory, security

- Determinism boundary - enforce separation between AI-assisted interpretation and deterministic computation

- Multi-posture deployment - ensure the system runs on-premises, hybrid, and cloud with identical trust guarantees

- Knowledge engineering - design how domain knowledge is captured, versioned, and consumed by both AI and deterministic components

- Traceability and governance - audit trails, human-in-the-loop approval gates, structured error handling

- Portfolio scaling - generalize patterns across agent use cases; maintain architecture documentation.

Requirements :

- Education: Bachelors or masters equivalent.

- 7+ years software architecture, 3+ years involving AI/ML systems in production

- Experience separating deterministic and probabilistic concerns in a single system

- Multi-environment deployment where data sovereignty or offline operation were real constraints

- Work in a regulated or safety-adjacent domain (energy, medical, automotive, aerospace, industrial)

- Generative AI Agentic orchestration frameworks, RAG, designing LLMs as orchestrators rather than decisionmakers

- Safety-critical design: Trust boundaries between probabilistic and deterministic components, audit and provenance architecture

- Knowledge engineering: Structured and versioned domain knowledge, knowledge graphs or semantic technologies

- Deployment: multi-posture delivery (on-prem + cloud), local and cloud-hosted LLM inference, CI/CD

- Security & identity: Auth2/OIDC, zero-trust principles, least-privilege access for AI tool integration

- AI evaluation: prompt/model versioning, regression testing, evaluation pipelines, model promotion governance

- Integration design: API design across heterogeneous systems, complex domain data format parsing and transformation

- Observability: Monitoring, tracing, and logging for production AI systems

- Architecture practice: structured documentation, decision records, hypothesis-driven design

Desired Domain Exposure :

- Electrical power systems or protection engineering

- Industrial automation, OT/IT convergence

- Energy sector product development

Technology Landscape :

- Python, TypeScript, LLM orchestration frameworks, Knowledge graphs, Azure cloud, Local LLM inference, OAuth2/OIDC, IEC 61850, COMTRADE.