Obed Kwesi Asamoah A.

Key Skills

Software

Label Studio

CVAT

Argilla

Top Subject Matter

Software Engineering & API Architectures

E-Commerce & Retail Marketplace Logistics

Computer Vision & Intelligent Systems

Top Data Types

Image

Text

Document

Top Task Types

Computer Programming/Coding

Function Calling

RLHF

Prompt + Response Writing (SFT)

Evaluation/Rating

Classification

Bounding Box

Freelancer Overview

Founder and Chief Executive Officer in Contract Review, Compliance, and Legal Research. Brings 12+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Education includes Bachelor of Science, Valley View University (2024). Well suited for text-focused AI training, including legal document review, compliance annotation, and rubric-based quality evaluation.

Labeling Experience

LLM Fine-Tuning & Sentiment Optimization for Global Digital Brands

TextEntity (NER) Classification

Scope: Created high-quality training matrices to fine-tune Large Language Models (LLMs) used in automated customer service, digital campaign analytics, and localized product copy generation across multi-regional distribution networks. Tasks Performed: - Conducted exhaustive Named Entity Recognition (NER) across thousands of raw, unstructured customer touchpoints, tagging brands, localized currencies, order IDs, and shipping terminology. - Developed complex classification maps for multi-tiered sentiment analysis (Positive, Mixed, Critical-Escalated) to flag high-churn consumer complaints automatically. - Generated high-quality instruction-tuning pairs (Prompt-Response datasets) to guide conversational conversational AI in alignment with brand voice guidelines. Project Size: Managed annotation, curation, and structured formatting of 40,000+ unstructured text entries and customer review datasets

2026 - Present

Verbatim Audio Transcription for Voice Interface Training

AudioTranscription

Scope: Converted raw multi-speaker voice notes and audio clips into clean, highly readable text datasets to fine-tune automated speech-to-text engines. Tasks Performed: - Listened to short audio snippets and typed out completely verbatim textual transcripts, noting exact vocal fillers (e.g., "uh", "um"). - Placed proper speaker differentiation tags (e.g., Speaker 1, Speaker 2) across multi-party files. - Used basic style guides to handle background noise interruptions or cut-off sentences cleanly. Project Size: Transcribed over 150 hours of short-form audio clips and voice recordings.

2025 - Present

LLM Model Output Evaluation and Human Preference Rating

TextEvaluation/Rating

Scope: Reviewed, benchmarked, and graded side-by-side text outputs generated by Large Language Models (LLMs) to optimize overall helpfulness, clarity, and factual alignment. Tasks Performed: - Read paired model variations responding to single consumer prompts and explicitly rated them on clear scales (e.g., accuracy, tone, format adherence). - Flagged bad generations containing hallucinated details, grammatical errors, or broken markdown formatting. - Wrote short, clear justifications explaining exactly why one model output was superior to the alternative. Project Size: Evaluated and assigned preference ratings to over 2,500 conversational text output pairs.

2025 - 2026

2D Bounding Box Object Annotation for Retail Inventory Systems

ImageBounding Box

Scope: Applied precise 2D bounding boxes over consumer packaged items and retail store inventory imagery to train object detection models for smart retail checkout systems. Tasks Performed: - Drew tight, precise rectangular bounding boxes around individual items on retail shelves, ensuring no accidental overlap or clipped borders. - Labeled specific target classes (e.g., beverages, snacks, hygiene items) according to strict object-boundary instructions. - Remade or adjusted loose, inaccurate bounding boxes submitted in baseline raw datasets. Project Size: Annotated 8,500 distinct product images across varied lighting conditions. Quality Measures: Achieved a 98.8% tight-margin accuracy score. Successfully passed internal quality loops with zero tolerance for missed objects or loose borders.

2025 - 2026

Open-Source Data Collection and Curation for Local Consumer Trends

TextData Collection

Scope: Sourced, collected, and neatly organized high-quality public domain text entries, public consumer reviews, and open-source imagery to seed a foundational marketing analytics database. Tasks Performed: - Gathered specific real-world text reviews, promotional materials, and product descriptions from public digital storefronts based on precise client briefs. - Standardized raw scraped web information into clean, readable CSV files with unified text formats. - Verified all collected links and media files were completely free of layout corruption or missing values. Project Size: Collected, compiled, and structured a baseline dataset of 6,500 distinct consumer data points.

2025 - 2025

Show more

Education

Work History