You will build and maintain AI-driven document processing pipelines, focusing on RAG architectures and LLM-integrated applications.
This role is remote.
Responsibilities
- Build document processing pipelines implementing OCR extraction, text normalization, and structured output generation for scanned and handwritten documents.
- Develop vector database layers including index schemas, embedding computation, metadata storage, and retrieval optimization using k-NN search.
- Implement AI classification engines in Python that retrieve domain knowledge from vector stores and execute multi-label classification with confidence scoring.
- Build REST API ingestion layers and RPA integrations to ingest documents from enterprise content management systems.
- Develop human-in-the-loop validation interfaces for SME review, logging all decisions for audit and tuning confidence thresholds.
Required Skills
- 7+ years in software development with 2+ years on AI/ML, NLP, or document processing solutions.
- Advanced Python: FastAPI/Flask for API development, pandas/numpy for data processing.
- LLM APIs: Azure OpenAI (GPT-4o/GPT-4), Anthropic Claude, or equivalent for text classification and embedding generation.
- RAG Orchestration: Hands-on experience with LangChain or LlamaIndex.
- Vector Databases: Production experience with OpenSearch, Pinecone, Weaviate, or Chroma for index creation and query optimization.
- OCR/Document Processing: Azure AI Document Intelligence, Tesseract, or ABBYY for scanned and handwritten text.
- Low-code platforms: Power Apps, Palantir Foundry, or Retool for building HITL review interfaces.
- RPA: UiPath Studio for ingestion automations and UiPath Orchestrator for job scheduling.
- DevOps: Git, CI/CD pipelines via Azure DevOps or UiPath Automation Ops.
Preferred Skills
- Domain experience in regulated industries (healthcare, insurance, financial services).
- Strong debugging skills across OCR, embeddings, LLM responses, and application logic.
- Ability to write clear deployment guides, API specifications, and technical documentation.