Sajeed A.

Key Skills

Software

No software listed

Top Subject Matter

Mathematics

Physics

Chemistry

Top Data Types

Image

Text

Document

Top Task Types

Classification

Text Summarization

Data Collection

Evaluation/Rating

Prompt + Response Writing (SFT)

Question Answering

Freelancer Overview

I build production pipelines that turn scanned and photographed documents into clean, structured, machine-ready datasets. Working hands-on with large language models (Gemini Vision and Claude), I designed an end-to-end extraction system that converts figure-heavy exam PDFs into verified structured JSON, with quality-assurance logic for duplicate detection via content hashing, answer-key quarantine for ambiguous items, and routing of figure-bearing questions to a separate human-review queue — achieving zero fabricated entries across validation batches. I ran a model bake-off to select the most accurate extraction model, scoring each output for correctness and hallucination before production use. My work is strongly multilingual: I extract and verify mixed-script source material in Bangla, Arabic, Urdu, and English, including documents where standard OCR fails and tabular content needs cell-level checking. My domain depth is in STEM — Mathematics, Physics, Chemistry, and Biology — built through authoring and verifying large question banks against source. Core strengths: data accuracy, document and text annotation, structured data labeling, LLM output evaluation, and rigorous validation.

Education

Work History