Forward Deployed AI/ML Engineer IV

We need to output HTML fragment with sections: About the role? Actually the posting includes description. We need headings: About the role, Responsibilities, Requirements, Qualifications, Skills, Benefits, Pay, Schedule. Only include sections with real content. We must not repeat job title as heading. No Job ID etc. Remove boilerplate like EEO, legal, application instructions, cookie/navigation etc. We need to preserve all substantive info. Let's parse content: Intro: At E Source, we help utilities... (company description). That could be "About the role"? Actually "About the role" maybe includes description of role and company. Could put under

About the role

with paragraphs. Then responsibilities: list of duties: embed with utility clients to design, build, deploy GenAI/ML systems; own technical architecture; run technical discovery; deliver working system early; design task-specific evaluations; build retrieval and grounding; integrate with client data platforms; instrument systems; build operator-facing application layer; set realistic expectations; manage scope; contribute reusable accelerators; apply AI safety etc. We can put those as

• items under Responsibilities. Then "You’re Likely a Great Fit If You Are ..." that's qualifications/skills. Could be Requirements or Qualifications. Also "And Even Better If You ..." additional preferred qualifications. Then "Experience And Skills To Qualify Include" list of required experience and skills. Then benefits: "What You Can Expect Excellent insurance options..." includes benefits: insurance, FTO, 401k match, salary range, remote-based with travel, work auth, etc. Also compensation range. We need Benefits section: list benefits. Pay section: salary range. Schedule: remote-based with significant travel 30-50% etc. We must not include filler like "commensurate with experience". The posting includes "Actual pay will be adjusted based on experience." That's okay. We must not include duplicated lines. We need to ensure we don't include boilerplate like EEO, visa sponsorship, etc. Those are boilerplate; we should remove. Also remove "Learn more at www.esource.com" maybe not needed but it's substantive? It's a link; could be considered minor metadata; maybe drop. We must not include application instructions like "To foster a positive work environment..." etc. That's boilerplate. Let's craft HTML. Structure:

  About the role

  At E Source, we help utilities make sense of complexity in a rapidly changing landscape... (company description). As a Forward Deployed Engineer, AI/ML, you’ll join our AI and Data Engineering (AIDE) team and embed directly with our utility clients to understand their hardest operational problems, then build and ship the systems that solve them: production code, in the client’s environment, used by their people, and measured against their outcomes. This role focuses on taking AI from a promising demo to a system a utility’s business runs on: grounded, evaluated, monitored, and trusted by the operators who depend on it. You’ll own engagements end to end, from technical discovery through architecture, build, deployment, and adoption, with no handoff points in between. Engagements move quickly: expect a useful system running in the client’s environment within weeks, then hardened toward production based on how their teams actually use it. You’ll work closely with client technical teams and executives, and with E Source’s machine learning engineers, data engineers, software engineers, and consultants, to deliver AI solutions for clients and bring field learnings back into our products and practices.

  Then responsibilities heading.

  Responsibilities

  • Embed with utility clients to design, build, and deploy generative AI (GenAI) and machine learning (ML) systems that solve real operational problems in the client’s environment
  • Own the technical architecture of each engagement — retrieval, orchestration, model selection, serving, evaluation, and monitoring — and defend it to client architects and security teams
  • Run technical discovery independently, separate the problem as stated from the underlying need, and say so when the requested solution is the wrong one
  • Deliver a working, useful system early in each engagement, typically within the first few weeks, then iterate toward production hardening — tactical solutions are legitimate, but undocumented or unowned ones are not
  • Design task-specific evaluations before building, define what “correct” means with client subject matter experts, and hold every system to those evaluations before release
  • Build retrieval and grounding systems over client data, including structured, unstructured, and graph-backed knowledge sources
  • Integrate with client data platforms and business systems, including undocumented and legacy systems
  • Instrument systems so accuracy, latency, cost, and drift are measured rather than asserted, and hand off operations the client’s team can run without us
  • Build and deploy the operator-facing application layer — review and approval interfaces, agent and chat front ends, evaluation dashboards, and workflow tools — so delivered systems are usable by client teams, not just their engineers
  • Set realistic expectations with executives about what AI can and cannot do for their problem
  • Manage scope, timeline, and expectations against a defined statement of work, and raise risk early
  • Contribute reusable accelerators, evaluation harnesses, and reference architectures that scale across clients, and share implementation feedback with E Source’s engineering and product teams
  • Apply AI safety, data privacy, and governance controls appropriate to the utility regulatory context
  Then maybe "Qualifications" heading for required experience and skills. But we also have "You’re Likely a Great Fit If You Are ..." and "And Even Better If You ..." and "Experience And Skills To Qualify Include". Could combine into Requirements and Skills sections. We need to decide: Use headings Requirements, Qualifications, Skills. The instruction: use headings for section headings (e.g. About the role, Responsibilities, Requirements, Qualifications, Skills, Benefits, Pay, Schedule). So we can have Requirements, Qualifications, Skills. We need to map content: - Requirements: maybe the "You’re Likely a Great Fit If You Are ..." (core fit) and "Experience And Skills To Qualify Include" (required). Could treat as Requirements. - Qualifications: maybe the "And Even Better If You ..." (preferred). - Skills: maybe the specific technical skills list (Python, etc.) but that's already in Experience And Skills. We could split: Requirements = core fit + required experience; Qualifications = preferred nice-to-have; Skills = maybe not needed separate. But we need to avoid thin sections. Perhaps combine Requirements and Qualifications into one? But instruction says prefer few substantive sections over many thin ones. We could have Requirements (includes both required and preferred) and Skills (technical skills). But we already have technical skills in Experience And Skills. Let's think: The posting includes: - "You’re Likely a Great Fit If You Are ..." (list of soft/behavioral fit) - "And Even Better If You Hold a master’s degree or PhD ..." (preferred qualifications) - "Experience And Skills To Qualify Include" (list of required experience and skills) Thus we could have:

  Requirements

  containing the "Experience And Skills To Qualify Include" list (required).

  Qualifications

  containing the "And Even Better If You ..." list (preferred).

  Skills

  maybe not needed; but we could put the "You’re Likely a Great Fit If You Are ..." as Skills? Actually those are more about abilities/behaviors; could be Skills. But we already have Responsibilities. Could treat "You’re Likely a Great Fit If You Are ..." as Skills (behavioral/soft skills). Then Qualifications for preferred education/experience. Then Requirements for required experience and skills. Let's do that. Thus:

  Skills

  • Comfortable owning an engagement end to end — from discovery through architecture, build, deployment, and post-go-live support — rather than a single phase of it
  • Navigate ambiguity, incomplete data, and objectives that shift mid-engagement without losing traction
  • Communicate clearly in writing and can whiteboard your thinking for both engineers and executives
  • Are candid about what didn’t work, why, and what you changed in your practice because of it
  • Move fast without cutting corners, are comfortable delivering a working system in weeks rather than quarters, and know what to defer
  • Work well with enterprise clients and can navigate stakeholders ranging from individual contributors to executives

  Qualifications

  • Hold a master’s degree or PhD in a relevant quantitative field
  • Have built agent frameworks and tool-calling architectures in production
  • Bring knowledge graph or semantic layer experience for grounding and reasoning
  • Have fine-tuning, distillation, or prompt optimization experience — and the judgment to know when they’re not the answer
  • Bring classical ML experience (forecasting, ranking, anomaly detection) and know when to choose it over a large language model (LLM)
  • Are fluent in data engineering: Spark, pipeline design, and lakehouse architectures
  • Have built and deployed Databricks Apps, or comparable operator-facing applications on a modern frontend framework (React, Streamlit, Dash, or Gradio) with REST or GraphQL APIs
  • Hold Databricks or cloud AI/ML certifications

  Requirements

  • Bachelor’s degree in computer science, engineering, statistics, or a related quantitative field
  • Five or more years of engineering experience, including generative AI (GenAI) or machine learning (ML) systems shipped to production and iterated on based on real usage
  • Expert proficiency in Python and working proficiency in at least one of TypeScript, Scala, or Java
  • Hands-on production experience with at least one major model provider or open-weights stack, and one vector or hybrid retrieval system
  • Experience building and applying evaluation frameworks for large language model (LLM) and ML systems, including retrieval-augmented generation (RAG), agentic or multi-agent systems, structured extraction, or text-to-SQL
  • Experience with LLM operations (LLMOps): model serving, latency and cost management, prompt and model versioning, monitoring, and regression detection
  • Working knowledge of at least one major cloud ecosystem (Amazon Web Services (AWS), Azure, or Google Cloud Platform (GCP)); experience with AWS and Databricks strongly preferred
  • Proficiency with Git, Docker, and continuous integration and continuous delivery (CI/CD) tools
  • Experience using AI coding assistants and agentic development tools to rapidly prototype, test, and iterate on working software
  • Experience in the energy or utility industry, or another asset-heavy sector, with working familiarity of the systems, data, and operational constraints utilities face
  • Experience shipping on compressed timelines, from first conversation to a working system in weeks rather than quarters
  Now Benefits section: From "What You Can Expect Excellent insurance options, including medical, dental, and vision plans; company-paid life insurance; company-paid long- and short-term disability insurance; medical and dependent-care flexible spending plans, and paid parental leave.A flexible time off (FTO) policy that provides paid time away from work, approved by your manager, while ensuring business needs, workload commitments, and appropriate coverage are maintained.A 401(k) plan with a 3% employer match." We can list as
  . Also note remote-based with travel etc. That's Schedule. Pay section: salary range. Schedule: remote-based with significant travel 30-50% etc. We must not include boilerplate like visa sponsorship, EEO, etc. Thus:

  Benefits

  • Excellent insurance