Senior AI Engineer – Agentic AI (Data Center Energy)

Elpho; Atlanta (GA); Hybrid; USD 150,000 - 210,000; Full time

Job summary

Elpho seeks a senior software engineer who has transitioned into AI and shipped AI/ML systems to production for energy data centers. You will build AI agents that operate with live power and cooling equipment, ensure safe guardrails, and integrate with Azure OpenAI and AKS.

Hybrid in Atlanta with remote options across the US. You will own alarm handling, forecasting, and knowledge retrieval, collaborating with field teams to improve models and deliver reliable energy optimization across BESS,

Qualifications

• 8–10+ years of professional software engineering (backend or distributed systems).
• 2–3+ years building AI/ML that shipped to production.
• Strong Python, plus production experience in at least one backend language (C#/.NET, Go, Java, or TypeScript).
• LLM application engineering: agents and tool calling, retrieval-augmented generation, context design, and structured outputs.
• Classical ML and time-series: forecasting, anomaly detection, and optimization.
• Evaluation discipline: measure model and agent quality before and after release.

Responsibilities

• Alarm and operations agents — AI agents that triage alarm and event streams, correlate them with live telemetry and the site’s digital twin, and recommend or execute responses inside approved guardrails, with human-in-the-loop approval.
• Optimization and forecasting — load and price forecasting, BESS dispatch, cooling and HVAC setpoint optimization, demand-charge management, and anomaly detection on high-frequency time-series data
• Knowledge and retrieval — retrieval-augmented answers grounded in SOPs, procedure libraries, equipment manuals, point lists, and the site knowledge graph
• Guardrails and safety — deterministic policy checks, action limits, approval workflows, and full audit trails
• Evaluation — offline evaluation harnesses, backtesting against historical site data, LLM quality metrics, and cost and latency budgets
• Production engineering — AI services on Azure (Azure OpenAI, AKS or Functions, Event Hubs) exposed through APIs
• Field feedback — use operators' acceptance to improve models and explain model behavior to operators and customers

Skills

Python

Backend development

AI/ML production

LLM application engineering

Time-series forecasting

Model evaluation

Azure cloud services

Safety and reliability

Tools

Azure OpenAI

LangGraph

Semantic Kernel

scikit-learn

PyTorch

Azure ML

AKS / Event Hubs

Job description

Our client is building the energy intelligence layer for AI data centers: a platform that monitors and manages a facility’s power systems — battery storage (BESS), switchgear, UPS, distribution, and SCADA/edge — optimizes energy across battery, grid, and HVAC load, and measures and settles every dollar saved in a daily ledger. AI agents handle alarm response and day-to-day operations. The first production data center is going live now, with a multi-site fleet rollout behind it. We are hiring the senior engineer who builds that AI layer: agents and models that operators trust enough to approve, running against live power and cooling equipment where a wrong action has real consequences. This role is for a seasoned software engineer who moved into AI and has shipped it to production. Atlanta preferred (hybrid); remote anywhere in the US considered.

What you'll do

• Alarm and operations agents — AI agents that triage alarm and event streams, correlate them with live telemetry and the site’s digital twin, and recommend or execute responses inside approved guardrails, with human-in-the-loop approval
• Optimization and forecasting — load and price forecasting, BESS dispatch, cooling and HVAC setpoint optimization, demand-charge management, and anomaly detection on high-frequency time-series data
• Knowledge and retrieval — retrieval-augmented answers grounded in SOPs, procedure libraries, equipment manuals, point lists, and the site knowledge graph
• Guardrails and safety — deterministic policy checks, action limits, approval workflows, and full audit trails, so the AI never takes an unsafe action on live equipment
• Evaluation — offline evaluation harnesses, backtesting against historical site data, LLM quality metrics (accuracy, grounding, hallucination rate), and cost and latency budgets
• Production engineering — AI services on Azure (Azure OpenAI, AKS or Functions, Event Hubs) exposed through APIs the platform and UI consume, with CI/CD, monitoring, and drift detection
• Field feedback — use what operators accept and reject at production sites to improve the models and agents, and explain model behavior plainly to operators and customers

What we're looking for

• 8–10+ years of professional software engineering (backend or distributed systems), followed by 2–3+ years building AI/ML that shipped to production — not just notebooks or prototypes
• Strong Python, plus production experience in at least one backend language (C#/.NET, Go, Java, or TypeScript)
• LLM application engineering: agents and tool calling, retrieval-augmented generation, context design, and structured outputs, using model APIs such as Azure OpenAI or Anthropic and frameworks such as LangGraph or Semantic Kernel (or your own)
• Classical ML and time-series: forecasting, anomaly detection, and optimization (scikit-learn, PyTorch, statsmodels; OR-Tools or Pyomo a plus)
• Evaluation discipline: you measure model and agent quality before and after release, and you can show how
• Cloud ML on Azure (Azure ML, Azure OpenAI, AKS, Event Hubs) or the AWS/GCP equivalents, and experience with streaming and time-series data
• Sound judgment about safety and reliability in systems that touch physical equipment

Nice to have

• Energy, power systems, building management (BMS), or SCADA experience; battery storage, demand response, or grid services
• Knowledge graphs and semantic models (Neo4j, Cosmos DB Gremlin, Brick or Project Haystack)
• MLOps tooling (MLflow, Azure ML pipelines) and edge or on-premise inference
• Experience in an early-stage product team moving fast with senior visibility