Oluwagbemiga Williams

Key Skills

Software

Other

Mindrift

Top Subject Matter

Frontier LLM Safety

Multilingual Adversarial Prompting

Trust & Safety

Top Data Types

Text

Document

Top Task Types

Red Teaming

Prompt + Response Writing (SFT)

Freelancer Overview

AI Adversarial Prompt Specialist — French Language. Brings 8+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Other and Mindrift. Education includes Master of Business Administration, Bayero University, Kano (2021) and Bachelor of Science, Oduduwa University, Ile-Ife (2017). AI-training focus includes data types such as Text and labeling workflows including Red Teaming, Evaluation, and Rating.

Labeling Experience

AI Adversarial Prompt Specialist — French Language

OtherTextRed Teaming

Designed and executed French-language adversarial prompts to test and probe the safety guardrails of frontier AI LLMs for Alice.io. Crafted structured scenarios targeting unsafe content generation, guardrail bypass, policy compliance boundaries, and documented findings for the safety research team. Applied narrative and indirect elicitation techniques across content restrictions and policy edge cases to expose model vulnerabilities. • Developed adversarial prompts to elicit unsafe or non-compliant outputs • Operated under strict ethical guidelines, ensuring vulnerabilities were revealed without promoting harm • Provided a multilingual perspective critical for cross-lingual testing and identifying policy gaps • Collaborated with researchers to document failure modes and reproduction steps

Designed and executed French-language adversarial prompts to test and probe the safety guardrails of frontier AI LLMs for Alice.io. Crafted structured scenarios targeting unsafe content generation, guardrail bypass, policy compliance boundaries, and documented findings for the safety research team. Applied narrative and indirect elicitation techniques across content restrictions and policy edge cases to expose model vulnerabilities. • Developed adversarial prompts to elicit unsafe or non-compliant outputs • Operated under strict ethical guidelines, ensuring vulnerabilities were revealed without promoting harm • Provided a multilingual perspective critical for cross-lingual testing and identifying policy gaps • Collaborated with researchers to document failure modes and reproduction steps

2025 - Present

Founder & Lead — AI Community

OtherText

Led an AI community focused on practical education, testing, and evaluation of AI tools in multilingual, low-resource African contexts. Conducted ongoing evaluation and surfacing of AI model failure modes, particularly regarding unexpected outputs and cross-lingual issues. Provided real-world insights valuable for multilingual red teaming and bias detection. • Educated members on LLM prompt engineering and responsible usage • Regularly tested and documented AI model behaviors across cultures • Surfaced bias and coverage gaps in mainstream AI systems • Shared findings for community-driven improvement of AI tools

Led an AI community focused on practical education, testing, and evaluation of AI tools in multilingual, low-resource African contexts. Conducted ongoing evaluation and surfacing of AI model failure modes, particularly regarding unexpected outputs and cross-lingual issues. Provided real-world insights valuable for multilingual red teaming and bias detection. • Educated members on LLM prompt engineering and responsible usage • Regularly tested and documented AI model behaviors across cultures • Surfaced bias and coverage gaps in mainstream AI systems • Shared findings for community-driven improvement of AI tools

2023 - Present

Head of Content / AI Workflow Architect

OtherTextPrompt Response Writing SFT

Developed a 13-column prompt engineering system and 15-point QA workflow for pre-publication evaluation of AI-generated text content. Evaluated outputs across multiple LLM architectures, maintaining strict quality control and compliance standards through human-in-the-loop review. Built governance systems to reduce errors and ensure only safe, accurate outputs passed to publication. • Engineered structured prompt libraries for Claude, ChatGPT, and Jasper • Maintained monthly production of 150+ evaluated AI content assets • Applied compliance logic and trust signals assessment to prompt reviews • Established pass/fail gates to catch unsafe or flawed outputs prior to release

Developed a 13-column prompt engineering system and 15-point QA workflow for pre-publication evaluation of AI-generated text content. Evaluated outputs across multiple LLM architectures, maintaining strict quality control and compliance standards through human-in-the-loop review. Built governance systems to reduce errors and ensure only safe, accurate outputs passed to publication. • Engineered structured prompt libraries for Claude, ChatGPT, and Jasper • Maintained monthly production of 150+ evaluated AI content assets • Applied compliance logic and trust signals assessment to prompt reviews • Established pass/fail gates to catch unsafe or flawed outputs prior to release

2021 - Present

Content Strategist & AI Operations Lead

MindriftText

Led AI model training operations for tech and fintech clients by evaluating LLM outputs for quality, factual accuracy, intent alignment, and compliance. Built structured AI content QA frameworks and developed hybrid AI-human evaluation workflows. Managed evaluators and documented output failures to contribute to model improvement cycles. • Designed structured criteria for LLM output assessment • Flagged and reported hallucinations, misalignment, and compliance issues • Standardized benchmarks for AI-assisted content in SaaS and fintech • Supervised a team to ensure consistent ratings across content

Led AI model training operations for tech and fintech clients by evaluating LLM outputs for quality, factual accuracy, intent alignment, and compliance. Built structured AI content QA frameworks and developed hybrid AI-human evaluation workflows. Managed evaluators and documented output failures to contribute to model improvement cycles. • Designed structured criteria for LLM output assessment • Flagged and reported hallucinations, misalignment, and compliance issues • Standardized benchmarks for AI-assisted content in SaaS and fintech • Supervised a team to ensure consistent ratings across content

2024 - 2024

Content Manager — Regulated Multilingual

OtherText

Produced and evaluated regulated multilingual content, building compliance-first workflows with multi-layered review for legal and quality governance. Assessed output under FCA, SEC, BaFin, and aviation content frameworks to ensure zero compliance violations. Developed risk assessment instinct to identify content crossing legal or ethical boundaries before publication. • Managed reviewed outputs across 6 languages and regulatory marks • Integrated compliance checks into structured evaluation workflows • Ensured alignment with financial and industry regulations in AI-assisted content • Achieved 1,200+ pages evaluated with zero compliance incidents

Produced and evaluated regulated multilingual content, building compliance-first workflows with multi-layered review for legal and quality governance. Assessed output under FCA, SEC, BaFin, and aviation content frameworks to ensure zero compliance violations. Developed risk assessment instinct to identify content crossing legal or ethical boundaries before publication. • Managed reviewed outputs across 6 languages and regulatory marks • Integrated compliance checks into structured evaluation workflows • Ensured alignment with financial and industry regulations in AI-assisted content • Achieved 1,200+ pages evaluated with zero compliance incidents

2022 - 2023

Education

Work History