AI Systems Engineer

We are building a next-generation Human-in-the-Loop AI Automation Platform for the regulated life sciences and pharmaceutical industry. Our system ingests messy multi-department raw documents (R&D, Quality Control, Quality Assurance) and uses Intelligent Document Processing (IDP), Multi-Agent orchestration, and Fine-Tuned Open LLMs (Gemma / MedGemma) to automatically assemble regulatory submission dossiers (such as CTD Module 3.2.S, DMF, and ASMF).

We are looking for an ambitious, self-driven AI Systems Engineer to join as a founding member of our engineering team. You will work directly with leadership to architect, build, and deploy high-accuracy AI extraction and conversational feedback loops.

🔑 Key Responsibilities

1. Build IDP & Extraction Pipelines: Develop deterministic PDF and table parsing pipelines using tools like PyPDF, pdfplumber, Azure Document Intelligence, or Unstructured.io to extract complex data into Pydantic schemas .
2. Implement Agent Workflows: Build stateful, multi-agent conversational engines using LangGraph / PydanticAI to enable interactive Human-in-the-Loop (HITL) user feedback (e.g., merging split PDFs, re-mapping sections via chat).
3. Fine-Tuning & Model Integration: Fine-tune open-weight LLMs (Gemma 2, MedGemma) using QLoRA / Unsloth for structured JSON output and domain function calling.
4. Backend API Engineering: Design high-performance REST APIs using FastAPI and orchestrate async task queues for processing long-context multi-page regulatory documents.
5. Vector Search & State Persistence: Implement vector databases (Qdrant, PostgreSQL/pgvector, or Milvus) and persistent state checkpointers for multi-turn user memory.

🛠️ Key Skills & RequirementsMust-Have Skills (Technical Baseline)

• Python Mastery: Strong Python engineering skills (Async, Type hints, Dataclasses, Pydantic v2 ).
• LLM Engineering: Hands-on experience with LangGraph , LangChain , or PydanticAI for stateful agents, tool use, and function calling.
• IDP & PDF Parsing: Experience parsing unstructured data, OCR, and messy PDF layouts/tables.
• FastAPI: Ability to build and containerize ( Docker ) scalable RESTful endpoints.
• RAG & Vector Databases: Experience with vector stores (Qdrant, Milvus, ChromaDB, or pgvector) and hybrid search concepts.