Oladimeji Suraju

Key Skills

Software

AWS SageMaker

Google Cloud Vertex AI

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Computer Code Programming

Document

Image

Medical Dicom

Text

Video

Top Task Types

Action Recognition

Audio Recording

Bounding Box

Classification

Computer Programming Coding

Data Collection

Diagnosis

Emotion Recognition

Entity Ner Classification

Function Calling

Object Detection

Question Answering

Segmentation

Text Generation

Text Summarization

Transcription

Freelancer Overview

I am an experienced data scientist and AI training data expert with over five years of hands-on experience designing, labeling, and verifying high-quality datasets for machine learning and generative AI applications. My background spans the full data science pipeline, including data ingestion, cleaning, exploratory analysis, feature engineering, and the creation of deterministic, reproducible problem sets for AI model training. I have designed and annotated complex data science problems simulating real-world business scenarios across finance, telecom, e-commerce, and healthcare, ensuring each dataset and solution is rigorously validated for accuracy and reproducibility. My technical toolkit includes Python (pandas, numpy, scipy, scikit-learn, statsmodels), data visualization (matplotlib, seaborn), and advanced GenAI tools such as LLMs, RAG, prompt engineering, and vector databases. I am passionate about producing reliable, scalable training data that powers robust AI systems, and have integrated MLOps best practices and clear documentation to support model deployment and ongoing performance monitoring.

Labeling Experience

Data labeling for Health Care

Aws SagemakerImageBounding BoxClassification

The B2B AWS SageMaker Ground Truth project builds labeled medical image datasets to train ML models assisting radiologists in diagnosing cancer and cardiovascular conditions. Using SageMaker Ground Truth Plus, providers securely ingest de-identified images via S3, manage HIPAA-compliant labeling with expert workforces, and deliver annotated data for SageMaker training. Tasks include bounding boxes for abnormalities (e.g., tumors), entity extraction from reports, and video frame tracking. Scalable to 10,000–100,000+ images over 4–6 weeks, with automated pre-labeling cutting costs. Quality is ensured through multi-worker consensus, verification, active learning, and 95%+ accuracy targets, enabling safe, compliant diagnostic AI.

The B2B AWS SageMaker Ground Truth project builds labeled medical image datasets to train ML models assisting radiologists in diagnosing cancer and cardiovascular conditions. Using SageMaker Ground Truth Plus, providers securely ingest de-identified images via S3, manage HIPAA-compliant labeling with expert workforces, and deliver annotated data for SageMaker training. Tasks include bounding boxes for abnormalities (e.g., tumors), entity extraction from reports, and video frame tracking. Scalable to 10,000–100,000+ images over 4–6 weeks, with automated pre-labeling cutting costs. Quality is ensured through multi-worker consensus, verification, active learning, and 95%+ accuracy targets, enabling safe, compliant diagnostic AI.

2025 - 2025

Google Cloud Vertex for Businesses

Google Cloud Vertex AIImageClassificationObject Detection

The B2B Google Cloud Vertex AI project creates labeled datasets for manufacturing defect detection. Scope: ingest production images via Cloud Storage, manage expert labeling, enable end-to-end ML pipelines with active learning. Tasks: image classification, object detection, semantic segmentation of flaws. Quality: multi-worker consensus, verification, active learning, 95%+ accuracy targets

The B2B Google Cloud Vertex AI project creates labeled datasets for manufacturing defect detection. Scope: ingest production images via Cloud Storage, manage expert labeling, enable end-to-end ML pipelines with active learning. Tasks: image classification, object detection, semantic segmentation of flaws. Quality: multi-worker consensus, verification, active learning, 95%+ accuracy targets

2024 - 2024

Education

Work History