AI Engineer

About Viridium AI Viridium AI builds the material intelligence platform manufacturers use to find and eliminate hazardous materials — like PFAS — from their products, before regulators or customers do. Backed by Microsoft Azure, our platform unifies fragmented BOM, ERP and PLM data into a product-material knowledge graph, then layers science-constrained AI and a Chemical Digital Twin on top to flag risk across nested, multi-tier supply chains in minutes instead of months. We're past stealth — live on Azure Marketplace, and just co-launched ViTel with Tata Elxsi for medical device manufacturers. The team is ex-Oracle, Microsoft, Salesforce, Amazon and TCS, and we're still early enough that your code ships to real manufacturers within weeks.

The Role We need an AI Engineer who can take our platform's reasoning layer further — running model experiments, fine-tuning LLMs for our domain, and building custom agents that reason over materials, regulations, and supply-chain data. This isn't a chatbot wrapper role: agents you build will sit at the core of how the product flags chemical and material risk. You'll work directly with the founding engineering team, not through five layers of process.

What you will do: - Design, build, and deploy custom AI agents and multi-agent workflows using frameworks such as LangChain, LangGraph, LlamaIndex, AutoGen, or CrewAI - Build agentic workflows involving tool use, function calling, memory, retrieval-augmented generation (RAG), and multi-step reasoning over our product-material knowledge graph - Run model experiments — prompt engineering, evaluation, and benchmarking across open-weight and proprietary LLMs — to find what works for chemical and regulatory reasoning - Fine-tune language models (LoRA, QLoRA, PEFT, full fine-tuning) for domain-specific use cases on material and compliance data - Build and maintain evaluation pipelines to track agent/model accuracy, hallucination rate, latency, and cost - Integrate agents with our Neo4j knowledge graph, Postgres data, chemical intelligence, and external APIs - Collaborate directly with product, data science, and architects to turn ambiguous problems into shipped agentic features - Contribute to MLOps practices — model versioning, CI/CD for ML pipelines, monitoring — as the team scales this out - Write clean, production-grade, well-documented code and hold the line on quality in reviews - Run in Agile — sprint planning, stand-ups, retros — without needing to be told twice

What you will bring: - 4–5 years of experience in AI/ML engineering, with hands-on exposure to LLMs and generative AI in recent projects - Strong, demonstrable experience building custom Agents/Agentic AI systems — this is a must-have, not optional - Hands-on experience with LangChain / LangGraph or an equivalent framework (LlamaIndex, AutoGen, CrewAI, Semantic Kernel), including agent orchestration concepts like planning, tool-calling, and memory - Solid experience designing model experiments and evaluation frameworks — comparing prompts, models, and agent configurations - Practical experience fine-tuning LLMs (LoRA/QLoRA/PEFT, dataset prep, training loops, hyperparameter tuning) - Strong Python skills, with experience using ML/AI libraries (PyTorch, Hugging Face Transformers, etc.) - Experience working with vector databases (Pinecone, Weaviate, FAISS, Chroma, Milvus) and RAG pipelines - Good understanding of prompt engineering and LLM API integration (OpenAI, Anthropic, open-source models via Hugging Face/vLLM) - MLOps experience (model deployment/serving, CI/CD for ML, monitoring) is a strong plus, not a requirement - Can communicate clearly and work well in a small, high-context team

Why Viridium - Ship agents and models that are already reasoning over risk for global manufacturers and medtech companies — not a slide deck - Work on a genuinely hard, valuable problem (chemical/material risk) instead of another chatbot wrapper - Small team, high ownership — your decisions show up in the product next sprint - Direct access to founders and architects, not layers of middle management