Chelappan Kumar

• Designed and deployed a production-grade AI pipeline for image and document processing: ingesting user-submitted PDFs,
Excel sheets, scanned images, and photos with 98% extraction accuracy.
• Integrated YOLO-based object detection into document and visual inspection workflows, enabling automated identification of objects
and patterns across diverse real-world image formats.
• Built video and image processing pipelines using OpenCV for pre-processing, frame extraction, and feature detection to support
downstream AI inference.
• Developed domain-specific prompts and fine-tuned Llama-family LLMs for structured data extraction from complex visual and text
documents.
• Enabled processing of 5, 000+ documents/invoices weekly, reducing turnaround time from days to minutes through a scalable,
resilient workflow on AWS.
• Built OCR + RAG pipelines on complex medical PDFs and unstructured documents, enabling reliable Q&A and information retrieval.
• Operated as a forward-deployed engineer — collaborating directly with enterprise customer teams to translate product requirements
into working AI features and vision-based prompt strategies.
• Deployed and managed solutions on AWS/GCP for scalable LLM inference and pipeline orchestration with CI/CD best practices.

• Implemented LLM-driven NLP enrichment and classification using advanced prompt engineering to enhance medical entity
detection and improve research-grade data quality.
• Automated acquisition and standardization of 400, 000-500, 000+ pharma records per month, accelerating analytics turnaround and
downstream insight delivery.