Open to opportunities

Emmanuel Ihesi

@emmanuelihesi1

Message

I specialize in RLHF and multimodal AI evaluation, ensuring policy-compliant alignment and high-quality annotations.

Nigeria

Message

What I'm looking for

I’m looking for a team that values rigorous RLHF/LLM evaluation, rubric-based calibration, and policy-compliant multimodal quality auditing—where I can strengthen AI alignment and safety with multilingual, edge-case-focused judgment.

I’m an RLHF-certified AI Evaluation & Alignment Specialist with 3+ years of experience in multimodal AI quality auditing, LLM output evaluation, and alignment workflows. I’m recognized for high-judgment edge-case interpretation, rubric-based RLHF preference ranking, and maintaining evaluation consistency across complex text, image, audio, and video scenarios.

I calibrate and govern evaluation processes end-to-end—covering prompt failure detection, LLM response scoring, and SQE search quality evaluation—while supporting multilingual alignment across English, Igbo, and Nigerian Pidgin. With benchmarks like 97% accuracy across 500+ RLHF samples, 800+ hours of multilingual transcription QA, and 7,000+ image evaluations, I focus on producing actionable insights that strengthen AI safety and training data quality.

Experience

Work history, roles, and key accomplishments

Current

AI Video Annotation Auditor

Current

AtlasCapture.io / OneForma

Jan 2024 - Present (2 years 5 months)

Executed auditor-grade multimodal evaluation workflows on 300+ hours of autonomous-driving video datasets featuring occlusion, motion uncertainty, and multi-object tracking. Achieved auditor certification by calibrating and approving annotation batches and resolving instruction-following failures using rubric-based QA standards.

Multimodal AI Quality Auditing Object Tracking Assessment

Current

RLHF & AI Alignment Analyst

Current

Luel.ai / Welodata.io / Remotask

Jan 2023 - Present (3 years 5 months)

Performed multilingual multimodal evaluation across 7,000+ image datasets and evaluated 800+ hours of audio (English, Igbo, Nigerian Pidgin) to support conversational AI alignment and training. Maintained 97% consistency across 500+ RLHF preference-ranking and prompt-assessment tasks, including prompt failure detection and SQE workflows.

LLM Output Evaluation Multimodal Quality Auditing Prompt Failure Analysis