Machine Learning Engineer (6+ Years) - OCR Agentic Systems

**About the Role** We’re hiring a **Machine Learning Engineer \- OCR Agentic Systems** **with six\+ years experience** to build and own a next\-generation multi\-agent document processing platform. This is not a basic OCR integration role. You’ll design and develop a multi\-layer AI\-driven pipeline that processes complex business documents (POs, invoices, RFQs, shipping notices), extracts structured data, validates it against dynamic schemas, and delivers high\-accuracy outputs to downstream systems. A key focus is building intelligent self\-healing systems—especially a Recovery Agent that reprocesses only low\-confidence fields instead of sending entire documents for manual review. About the Role We’re hiring a Software Engineer \- OCR Agentic Systems to build and own a next\-generation multi\-agent document processing platform. This is not a basic OCR integration role. You’ll design and develop a multi\-layer AI\-driven pipeline that processes complex business documents (POs, invoices, RFQs, shipping notices), extracts structured data, validates it against dynamic schemas, and delivers high\-accuracy outputs to downstream systems. A key focus is building intelligent self\-healing systems—especially a Recovery Agent that reprocesses only low\-confidence fields instead of sending entire documents for manual review. **What You’ll Build** 🔹 **Multi\-Agent OCR Pipeline** You’ll design and implement a five\-layer agent system: * Format Detection Agent — identifies document type/format and routes processing * Primary Extraction Agent — orchestrates multiple OCR engines for optimal results * Document Classification Agent — classifies document types using AI models * Schema Matching Agent — maps extracted data to structured schemas with confidence scoring * Recovery Agent — re\-extracts low\-confidence fields using targeted AI processing 🔹 **Beyond the Pipeline** * MCP servers for agent\-tool communication * Confidence scoring engine with validation logic * Active learning feedback loops * OpenSearch\-based state tracking and observability * AWS\-based infrastructure (Lambda, ECS, SQS, Step Functions, Terraform) 🔹 **What You’ll Do** * Agent Architecture \& Orchestration * Build stateful agent workflows using LangGraph or similar * Design orchestration with AWS Step Functions * Implement multi\-engine OCR strategies (accuracy vs cost optimization) * AI \& Backend EngineeringDevelop LLM\-powered systems with structured outputs and tool usage * Build FastAPI services following clean architecture/DDD principles * Design efficient prompts and context flows * OCR \& Document ProcessingIntegrate and optimize OCR engines (Textract, DeepSeek\-OCR, vision models) * Implement image preprocessing (deskewing, denoising, enhancement) * Reliability \& ObservabilityImplement confidence\-based routing logic * Track pipeline execution using OpenSearch * Build monitoring dashboards and alerts * Data \& InfrastructureDesign schemas in PostgreSQL (pgVector) * Deploy scalable services using Docker, ECS/Fargate, Terraform * Build robust, well\-tested APIs 🔹 **Requirements** * Strong experience in Python and backend development * **Six\+ years experience** * Hands\-on experience building AI/LLM\-powered applications * Experience with agent frameworks (LangGraph, LangChain, etc.) * Solid understanding of OCR/document processing systems * Experience working with AWS services (Lambda, S3, ECS, Step Functions, etc.) * Strong API design and system design skills * Experience with PostgreSQL and search systems (OpenSearch/Elasticsearch) * Good testing practices, including AI system validation * Strong English communication skills (C1 preferred) 🔹 **Nice to Have** * Computer vision experience (OpenCV, Pillow) * Schema matching / fuzzy matching / semantic mapping * MCP or tool\-based LLM integrations * Human\-in\-the\-Loop (HITL) workflows * Experience in logistics, manufacturing, or document\-heavy domains * LLM observability tools (LangSmith, LangFuse, Arize Phoenix) * Experience with self\-hosted models or GPU workloads 🔹Why Join Us? * Work on real\-world agentic AI systems (not wrappers) * Build high\-impact systems improving accuracy from \~85% → 95%\+ * Own architecture and execution end\-to\-end * Collaborate directly with leadership and product teams * Solve complex, meaningful engineering challenges