Development and Optimization of AI Models: Maintain and evolve YOLOv8 models for document classification, BERT for normalization of medical exams, and integrate with GPT-4 for structured data extraction;
Serverless Architecture: Develop and optimize Azure Functions in Python, implementing asynchronous processing via Azure Service Bus with an event-driven architecture;
Intelligent OCR Pipeline: Enhance a pipeline that combines Azure Cognitive Services OCR with LLM-based post-processing, including checkbox detection, handwritten text processing, and medical forms handling;
Performance and Scalability: Optimize system throughput, implement intelligent batching, and reduce end-to-end latency;
Systems Integration: Maintain integrations with external APIs (CRM validation, Sensedia), PostgreSQL, AWS S3, and implement new connectors as needed.
Requirements
Strong experience with Python (3.9+) and development of distributed systems;
Hands-on experience with Azure (Functions, Service Bus, Blob Storage, Cognitive Services);
Experience deploying ML models to production (preferably computer vision and/or NLP);
Familiarity with frameworks: PyTorch/TensorFlow, Hugging Face Transformers, Ultralytics (YOLO);
Experience with document processing and OCR;
Knowledge of microservices architecture and asynchronous messaging.
**Nice-to-haves:**
Experience fine-tuning LLMs (e.g., GPT, BERT);
Familiarity with string-matching algorithms (Jaro-Winkler, fuzzy matching);
Experience with medical image processing or health-related documents;
Familiarity with observability tools (OpenTelemetry, Azure Monitor);