Buchi M.

Key Skills

Software

Don't disclose

OpenCV AI Kit (OAK)

Other

Top Subject Matter

AI/ML model evaluation and LLM output auditing for coding tasks

LLM fine-tuning data curation

CV annotation

Top Data Types

Image

Text

Document

Top Task Types

Segmentation

Bounding Box

Entity (NER) Classification

Freelancer Overview

AI & Code Quality Evaluator (Independent Consultant). Brings 6+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Don't disclose, Internal, and Proprietary Tooling. Education includes Master of Science, University of Hull and Bachelor of Science, IU University of Applied Science. AI-training focus includes data types such as Computer Code, Programming, and Image and labeling workflows including Evaluation, Rating, and Segmentation.

Labeling Experience

AI & Code Quality Evaluator (Independent Consultant)

Don't disclose

Serves as an expert evaluator for LLM-generated code, assessing correctness, architectural soundness, efficiency, and compliance with Python PEP 8. Designs nuanced multi-turn prompts and evaluates model reasoning through edge-case scenarios, debugging exercises, and mathematical logic tests. Produces structured critiques that isolate errors and improve output reliability for human-in-the-loop review and training use. • Vet and rank LLM code blocks for algorithmic and performance quality. • Author multi-turn prompt scenarios targeting reasoning and debugging behavior. • Perform fact-checking and logical/security fallacy isolation in model outputs. • Provide evaluation feedback suitable for training and alignment workflows.

2026 - Present

AI Data Curation Lead & Founder at Aricah

ImageSegmentation

Curates and engineers instructional datasets for fine-tuning open-source LLMs using parameter-efficient tuning (LoRA) with strict formatting rules. Designs and runs a human-in-the-loop visual annotation workflow for a kitchen assistant domain, validating food-image datasets for custom object recognition models. Audits LLM prompt-response outputs and supports red-teaming to mitigate hallucinations and improve response alignment in production deployments. • Curate and synthetically expand SFT-style instructional datasets for LoRA fine-tuning. • Annotate and validate food images for computer vision model training. • Benchmark and stress-test prompt-response datasets for production OpenAI and Gemini usage. • Build pipelines to filter, deduplicate, and vectorize corpora for RAG evaluation and semantic search.

2024 - 2026

Software Engineer (data validation/analytics focus) at Landmark Africa

Text

Performs automated data validation and structural alignment of large enterprise transaction datasets using scripted audits and batch routines. Maintains quantitative dashboards that track metrics across high-volume data streams to ensure integrity and error-free reporting. Supports downstream data quality processes that enable reliable model/data training inputs. • Write SQL scripts to clean, audit, and align transaction data structures. • Automate batch validation to reduce discrepancies in manual processing. • Manage analytics dashboards for metrics across millions of data points. • Monitor and ensure structural integrity and reporting accuracy.

2022 - 2024

African Languages NLP Initiative — Core Data Annotator & Developer

OtherTextEntity (NER) Classification

Acts as core data annotator and developer for a low-resource African languages NLP initiative. Handles data preprocessing, text alignment, and tokenization diagnostics to produce training-ready sequences. Performs grammatical sequence labeling to support training of a low-resource language processing application based on the open-source Waxal dataset. • Preprocess and align low-resource linguistic text corpora for training. • Conduct tokenization diagnostics and data quality checks. • Apply grammatical sequence labeling for model supervision. • Prepare structured outputs suitable for downstream NLP model training.

Not specified

Published Computer Vision Research (Zenodo) — Lead Researcher

OpenCV AI Kit (OAK)ImageBounding Box

Leads published computer vision research involving object detection and segmentation for autonomous driving datasets. Conducts extensive geometric data annotation, including bounding boxes and pixel-level semantic masking, across large-scale driving datasets. Compares model accuracy bounds across YOLOv8, Mask R-CNN, and DeepLabv3 using annotated ground truth. • Perform bounding-box labeling for autonomous-vehicle object detection tasks. • Create pixel-level segmentation masks for semantic segmentation evaluation. • Annotate large driving datasets to support comparative model benchmarking. • Evaluate detection/segmentation accuracy across multiple vision architectures.

Not specified

Education

Work History