Jima Dube

Key Skills

Software

AWS SageMaker

Dataloop

Top Subject Matter

Customer Support – Chatbots, Intent Classification & Sentiment Analysis

Finance – Risk Analysis, Fraud Detection & Financial Modeling

Cybersecurity – Threat Detection, Malware Analysis & Risk Assessment

Top Data Types

Document

Text

Top Task Types

Classification

Segmentation

Freelancer Overview

I have hands-on experience working with AI systems and data pipelines that directly support training data workflows, particularly in classification, data processing, and semantic search applications. At Shega Media and Technology, I developed an ML-powered classification system that predicts hierarchical categories from raw user input, improving tagging accuracy by over 85% and reducing manual labeling effort by 70%. This involved preparing and structuring datasets, cleaning and transforming data, and ensuring consistency for model performance. I also worked on large-scale data ingestion from multiple global sources (such as UN and FRED APIs), where I standardized and organized datasets into analytics-ready formats, aligning closely with high-quality training data requirements. In addition, I built AI-driven systems that leverage embeddings and vector databases (ChromaDB) to enable semantic search and real-time question answering over documents. This required generating high-quality text representations, optimizing data chunking strategies, and improving retrieval accuracy (achieving over 90% query precision). My experience combines data engineering, data preprocessing, and practical ML integration, allowing me to contribute effectively to AI training workflows, data labeling optimization, and quality assurance processes.

Labeling Experience

Education

Work History