Looking for a job

Ritu Sarsahay

@ritusarsahay

Machine Learning Engineer at Capsitech who replaced document extraction APIs with a fine-tuned model, raising accuracy to 95%.

India

Message

What I'm looking for

AI/ML Engineer with 3 years of hands-on experience building production AI solutions, specializing in Document Intelligence, GenAI, LLMs, RAG, and computer vision. Skilled in Python, PyTorch, VLM fine-tuning, and scalable AI pipelines, with a focus on improving accuracy, latency, and cost.

At Capsitech, I architected a document extraction system using PDFPlumber, PaddleOCR, and YOLOv9c to process digital and scanned PDFs with 92% field-level accuracy.

I also fine-tuned an LLM for email suggestions, increasing acceptance by 18% and reducing drafting time by 30%. A hybrid RAG system I designed improved retrieval relevance by 25% over keyword-only search.

On a Financial Data Extraction System project, I replaced Azure Document Intelligence with a pipeline combining YOLO-based field detection and a fine-tuned Qwen-VL model. It improved unknown-template extraction accuracy from 80% to 90% in internal evaluation.

I built an HMRC-style question-answering chatbot with hybrid retrieval and source-linked PDF querying. I also reduced representative scanned-document extraction time from 86 seconds to 20 seconds by moving inference to vLLM with request batching.

Experience

Work history, roles, and key accomplishments

Current

Machine Learning Engineer

Current

Capsitech

Jul 2024 - Present (2 years 3 months)

Architected an end-to-end document data extraction system processing 5,000+ documents/month with 92% field-level extraction accuracy. Improved email autocomplete/response quality by fine-tuning an LLM, lifting suggestion acceptance rate by 18% and reducing response drafting time by 30%. Designed and deployed a hybrid RAG system, improving retrieval relevance by 25%. Optimized model serving, increa