About AARC Environmental
AARC Environmental Inc. delivers market-leading EHS compliance solutions. We're embedding AI into our core products leveraging RAG, LangChain, and fine-tuned LLMs to automate regulatory tracking, risk assessment, and client reporting.
Job Overview
We are seeking a hands-on AI Engineer with deep experience in Generative AI, agentic AI, and ML engineering. You will architect end-to-end RAG systems and build production-grade AI agents that plan tasks, call tools/APIs, and autonomously orchestrate EHS workflows (e.g., document intake, permit monitoring, data entry/QA, risk flagging). You'll own model and agent design, evaluation, deployment, and monitoring in partnership with data engineers and compliance SMEs.
Responsibilities
Build Agentic AI Systems
- Design autonomous and human-in-the-loop AI agents that can plan, reason, and execute multi-step tasks (task decomposition, routing, reflection, retry/rollback).
- Implement tool-use/function-calling for agents (Power BI, Monday.com, JotForm, FileMaker Data API, SharePoint/OneDrive, Azure Functions, email/Teams, SQL/KQL endpoints).
- Develop multi-agent patterns (planner/solver/critic, researcher/writer/reviewer) using frameworks such as LangGraph, AutoGen, CrewAI, Semantic Kernel, or custom state machines.
- Add guardrails & policies (PII masking, allowlists, rate-limits, cost/latency budgets, escalation triggers).
Architect End-to-End RAG Pipelines
- Design vector-database architectures (e.g., FAISS) for sub-second retrieval over PDFs, permits, SOPs, inspection reports, and FileMaker exports.
- Build ingestion pipelines (chunking, embeddings, hybrid search, citations/grounding) and relevance feedback loops.
Lead LLM Fine-Tuning & Optimization
- Apply parameter-efficient finetuning methods (QLoRA, PEFT, adapter layers) to base models (e.g., LLaMA, Mistral, Claude) for EHS tasks (reg-summary, defect classification, corrective-action drafting).
- Prompt-engineering + tool-use orchestration with LangChain / custom routers; maintain prompt registries and evals.
Productionize, Observe, and Improve
- Ship agents/models behind secure APIs; implement versioning, canary, A/B, and automated regression tests.
- Set up agent telemetry (traces, spans, tool-latency, token/cost), drift/outlier alerts, and safety rails.
- Define eval suites (task success, factuality/grounding, latency, cost, user-satisfaction) and drive continuous improvement.
Collaborate & Evangelize
- Partner with BI engineers, data architects, and compliance experts to translate requirements into robust AI solutions.
- Document patterns and standards best practices for RAG, vector databases, and finetuning workflows.
- Lead knowledge-sharing sessions
Qualifications
- Bachelor's or Master's in Computer Science, AI/ML, or related field
- 5+ years in AI/ML engineering with 2+ years focused on RAG architectures and production AI agents.
- Strong Python proficiency and deep familiarity with PyTorch or TensorFlow; proficiency with LangChain or equivalent orchestration frameworks.
- Hands-on experience with vector databases (e.g., FAISS), LangChain, and RAG pipelines
- Proven use of function-calling/tool-use and API integrations (Monday.com, JotForm, FileMaker Data API, SharePoint/Graph, Power BI REST, Azure Functions).
- Proven track record of finetuning LLMs using QLoRA/PEFT/LoRA
- Hands-on with vector databases (FAISS/pgvector/milvus) and doc processing at scale (PDF parsing/OCR).
- Familiarity with automated testing and version control for model endpoints
- Clear communicator who can present complex AI concepts to non-technical stakeholders
- Familiarity with dashboarding platforms (Power BI)
- Exposure to RESTful API integrations (JotForm API, Monday.com)
