Taoheed H.

Key Skills

Software

No software listed

Top Subject Matter

Predictive analytics for student exam scores (ML model development and deployment)

Clinical hypertension risk prediction (MLOps pipeline development)

Binary diabetes classification (ML training and deployment pipeline)

Top Data Types

Computer Code Programming

Text

Document

Top Task Types

Function Calling

Diagnosis

Freelancer Overview

StudentIQ: End-to-End ML Exam Score Prediction System (For model-ready data preparation and ML training pipeline). Brings 6+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include ZenML, FastAPI, and Optuna. Education includes Diploma in Data Science, Appclick ICT Academy (2025) and Higher National Diploma in Insurance, Federal Polytechnic Offa (2024). AI-training focus includes data types such as Computer Code, Programming, and Medical and labeling workflows including Function Calling and Diagnosis.

Labeling Experience

StudentIQ: End-to-End ML Exam Score Prediction System (For model-ready data preparation and ML training pipeline)

Function Calling

Developed and deployed machine-learning exam score prediction services using an end-to-end pipeline approach focused on preparing model-ready training data. Used ZenML for orchestrating encoding, scaling, splitting, and model training, with Optuna hyperparameter tuning to improve regression performance. Exposed the trained model via a FastAPI REST API with validated request/response schemas for consistent inference. • Implemented ZenML training pipeline steps for data encoding, scaling, and train/test splitting. • Tuned a RandomForestRegressor using Optuna (30 trials) with 5-fold cross-validation to maximize R². • Deployed a FastAPI inference API using Pydantic schema validation for real-time predictions. • Added Logfire for real-time API observability and structured error monitoring.

2025 - Present

End-to-End Diabetes Prediction system (MLOps pipeline development)

Diagnosis

Created a production-style MLOps pipeline for diabetes risk classification, including orchestration, training, and deployment. Built modular, versioned workflow steps for repeatable experiments and artifact lineage tracking. Automated hyperparameter tuning and served predictions through a REST API with request/response validation and structured observability. • Implemented a modular ZenML pipeline with 5 versioned steps: loading, encoding, splitting, scaling, and training. • Ran Optuna hyperparameter optimization (30 trials, 5-fold stratified cross-validation) to improve F1 score. • Deployed a FastAPI inference service with Pydantic validation and Logfire-based structured logging. • Serialized and reloaded preprocessing artifacts (e.g., StandardScaler, LabelEncoder) to maintain training/serving consistency.

Not specified

End-to-end hypertension prediction system (ML pipeline and inference service)

Diagnosis

Built an end-to-end hypertension risk prediction system using a reproducible ML training pipeline and a live inference service. Created preprocessing and encoding steps that were applied consistently between training and serving to reduce data leakage. Implemented hyperparameter optimization, artifact versioning, and real-time prediction via a REST endpoint. • Designed a ZenML training pipeline with versioned ingestion, encoding, stratified splitting, feature scaling, and model training. • Optimized hyperparameters with Optuna (30 trials) using 5-fold stratified cross-validation to improve binary classification performance. • Deployed a FastAPI inference service with Pydantic validation and artifact retrieval from the ZenML model store. • Ensured consistent preprocessing between training and serving (label encoding + standard scaling) for reliable inference.

Not specified

Education

Work History