Python Developer & Machine Learning Engineer with ~4 years of professional experience (~46 months at RxLogix) building scalable microservices, intelligent document processing pipelines, and Agentic AI solutions for pharmacovigilance and healthcare workflows.

• 🛠 Core Expertise: Python, Django, FastAPI, Machine Learning, Deep Learning, NLP & Agentic AI Frameworks.
• 🩺 Domain Experience: Data Scientist & Software Engineer (ex-RxLogix), specialized in handling medical documentation, adverse event processing, and custom mapping solutions.
• ☁️ Infrastructure & Search: Experienced in AWS microservices, Docker, CI/CD pipelines, and high-performance querying with Oracle and OpenSearch.
• 🟢 Open Source: Active contributor to core Python document processing libraries (fpdf2, py-pdf).

Languages & Frameworks

AI, NLP & Search

Cloud, Databases & DevOps

• Engineered intelligent document processing pipelines to extract and validate structured, unstructured medical data and clinical records.
• Migrated legacy microservices to AWS infrastructure, improving system availability and processing speeds.
• Implemented Agentic AI frameworks and custom data mapping pipelines for pharmacovigilance operations.
• Worked with high-volume database engines including Oracle and OpenSearch for optimized search and retrieval.

I actively contribute to core Python document processing & PDF infrastructure:

• py-pdf/fpdf2: Contributed text parsing and structural rendering improvements (View my PRs).
• py-pdf/pypdf: Contributed document extraction & parsing enhancements (View my PRs).
• titipata/scipdf_parser: Enhancing scientific PDF parsing and section extraction pipelines (View my PRs).

• [Medical Safety Narrative Generator (NLG)]: Automated pipeline utilizing LLMs & fine-tuned transformer models to generate natural language safety reports from structured adverse event data.
• [Intelligent Document Processing (IDP) Engine]: Multi-modal OCR & parsing pipeline extracting structured JSON schemas from unstructured medical PDFs.
• [MedDRA Automated Event Coding Engine]: Built a hybrid search architecture combining dense vector embeddings (semantic search) and sparse lexical search (BM25/OpenSearch) to automatically map verbatim adverse event terms to standardized MedDRA codes with high precision.

• Stack: Python, Django, PostgreSQL, Celery, Redis
• Summary: Event-driven backend management platform handling asynchronous task queues, stock alerts, and real-time operational analytics.

• GitHub: @prateek-dagar