Mohammed B.

Key Skills

Software

AWS SageMaker

Label Studio

Other

Top Subject Matter

Legal/petition classification (Corrupt Practices Act 2000)

Legal reference chatbot (Corrupt Practices Act 2000)

Financial document analysis and fraud/compliance risk detection

Top Data Types

Text

Document

Image

Top Task Types

Classification

Question Answering

Data Collection

Object Detection

Fine-tuning

Text Summarization

Text Generation

Freelancer Overview

AI Petition Classifier (Project). Brings 8+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Don't disclose and Other. Education includes Bachelor of Technology, Abubakar Tafawa Balewa University (2021) and Secondary School Certificate Examination (SSCE), Nigerian Tulip International Colleges Yobe (2015). AI-training focus includes data types such as Text, Document, and Image and labeling workflows including Classification, Question Answering, and Entity (NER).

Labeling Experience

Skin Cancer Image Classification Using Machine Learning (Project)

OtherImageClassification

Trained and evaluated a machine learning image classification system for skin cancer using AlexNet and ResNet models. The project involved assigning labeled diagnostic classes to skin cancer image inputs to supervise model training and comparison. Model performance was assessed by comparing prediction accuracy across architectures. • Labeled skin cancer image classes for supervised training • Trained AlexNet and ResNet models on the labeled dataset • Compared architecture performance using accuracy metrics • Conducted evaluation to determine best-performing prediction approach

2024 - 2024

RegTech Compliance Monitoring Solution (Project)

Don't discloseText

Implemented automated compliance workflows for monitoring policy violations, including SDLC documentation review and audit reporting. The compliance solution required mapping policy rules to structured outputs and using labeled violation indicators to drive workflow triggers. It also supported regulatory monitoring by producing consistent, structured compliance findings from reviewed materials. • Defined compliance violation categories used for workflow outcomes • Used labeled examples to determine when policies were violated • Automated documentation review steps and audit reporting signals • Conducted monitoring assessments to ensure regulatory adherence

2024 - 2024

Azure AI Handwritten Document Evaluation Project (Project)

OtherDocument

Evaluated Azure AI Vision OCR capabilities on Nigerian handwritten documents to assess text extraction accuracy. The work focuses on measuring OCR performance for documents such as prescriptions, school forms, and receipts, which requires ground-truth text comparison. This implies creating or using labeled ground-truth transcription for evaluation and error analysis. • Prepared handwritten document samples for OCR testing • Compared OCR outputs against ground-truth transcriptions • Measured extraction accuracy and assessed common failure cases • Evaluated suitability for digital transformation based on OCR performance

2024 - 2024

AI Petition Classifier (Project)

Don't discloseTextClassification

Built and deployed an AI petition classifier to automatically categorize corruption-related petitions into legal offense and category labels. The model was designed to improve the speed and consistency of human review by producing structured classifications from petition text. This work involved preparing and using labeled legal text categories to train and evaluate the classifier. • Defined target label taxonomy for corruption-related offense categories • Prepared training data from petition texts with corresponding legal categories • Trained and validated a machine learning classifier for accurate categorization • Iterated model outputs to improve review-time efficiency and consistency

2022 - 2023

AI Petition Analyzer and Duplicate Detector (Project)

Don't discloseTextData Collection

Designed an AI-powered system to analyze petitions and detect duplicates using similarity scoring, recommending the most relevant existing case. Duplicate detection depends on assembling comparable petition texts and labeled relationships such as “duplicate/non-duplicate” or matched case pairs for model guidance. The output produced structured similarity-based recommendations for reviewing officers. • Built labeled or curated pairs for duplicate similarity evaluation • Generated similarity scores between new and existing petitions • Recommended the most relevant existing cases based on match strength • Tested and refined duplicate detection thresholds for better recall and precision

2019 - 2020

Show more

Education

Work History