For employers

Hire this AI Trainer

Sign in or create an account to invite AI Trainers to your job.

LoginCreate account
Invite to Job
V
Victory Ak

Victory Ak

AI Data Annotation & Segmentation – Robotic Arm Action Segmentation

United Kingdom flagLeeds, United Kingdom
$20.00/hrExpert

Key Skills

Software

No software listed

Top Subject Matter

Robotics Domain Expertise
Computer Vision
Speech Recognition

Top Data Types

VideoVideo
AudioAudio
TextText
ImageImage
DocumentDocument

Top Task Types

SegmentationSegmentation
TranscriptionTranscription
RLHFRLHF
Red TeamingRed Teaming
Fine-tuningFine-tuning

Freelancer Overview

AI Data Annotation & Segmentation – Robotic Arm Action Segmentation. Core strengths include Internal and Proprietary Tooling. AI-training focus includes data types such as Video, Audio, and Text and labeling workflows including Segmentation, Transcription, and Evaluation.

ExpertEnglish

Labeling Experience

Linguistics Expert – LLM Data Annotation & Refinement

TextFine Tuning
Used advanced linguistics expertise to train, evaluate, and refine large language models for natural language understanding and generation. Defined detailed annotation guidelines and assessment rubrics for LLM outputs. Supported LLM fine-tuning through structured evaluation and iterative data quality improvements. • Articulated best practices for linguistic data labeling and assessment • Evaluated and annotated LLM outputs for relevance and accuracy • Collaborated with engineering for model refinement • Consistently improved model natural language capabilities

Used advanced linguistics expertise to train, evaluate, and refine large language models for natural language understanding and generation. Defined detailed annotation guidelines and assessment rubrics for LLM outputs. Supported LLM fine-tuning through structured evaluation and iterative data quality improvements. • Articulated best practices for linguistic data labeling and assessment • Evaluated and annotated LLM outputs for relevance and accuracy • Collaborated with engineering for model refinement • Consistently improved model natural language capabilities

2026 - Present

Multi-Modal Action Timeline Annotation

VideoSegmentation
Created frame-accurate action segmentation and factual timelines for head-mounted camera videos demonstrating complex tasks. Omitted uncertain attributes and focused on visible subtask goals to maintain temporal and factual integrity. Monitored and enforced quality standards such as timestamp precision and factuality protocols. • Performed multi-modal segmentation of task action sequences • Rejected annotations for hallucinations or timing errors beyond threshold • Applied coarse terminology for ambiguous action directions • Maintained zero temporal gaps in timeline annotation

Created frame-accurate action segmentation and factual timelines for head-mounted camera videos demonstrating complex tasks. Omitted uncertain attributes and focused on visible subtask goals to maintain temporal and factual integrity. Monitored and enforced quality standards such as timestamp precision and factuality protocols. • Performed multi-modal segmentation of task action sequences • Rejected annotations for hallucinations or timing errors beyond threshold • Applied coarse terminology for ambiguous action directions • Maintained zero temporal gaps in timeline annotation

2026 - Present

Factuality Red-Teaming & Adversarial Prompting

TextRed Teaming
Designed and executed adversarial prompts to identify and document objective failures in large language models. Authored comprehensive Golden Responses and developed detailed atomic rubrics for model output evaluation. Evaluated outputs across diverse domains while maintaining rigorous quality standards. • Created standard, hard, and abstention-level adversarial prompts • Embedded logical contradictions and misinformation for testing • Evaluated model corrections to adversarial scenarios • Maintained 99% quality across multiple subject matter domains

Designed and executed adversarial prompts to identify and document objective failures in large language models. Authored comprehensive Golden Responses and developed detailed atomic rubrics for model output evaluation. Evaluated outputs across diverse domains while maintaining rigorous quality standards. • Created standard, hard, and abstention-level adversarial prompts • Embedded logical contradictions and misinformation for testing • Evaluated model corrections to adversarial scenarios • Maintained 99% quality across multiple subject matter domains

2026 - Present

Multi-Turn Audio RLHF & Steerability

AudioRLHF
Evaluated multi-turn voice interactions to enhance AI assistant conversational performance. Developed and applied rubric-based criteria to assess adherence to personas, accents, and delivery standards. Conducted transcription auditing and penalized model failures for robust reinforcement learning feedback. • Engaged in natural, multi-turn spoken dialogues • Applied atomic rubrics to assess conversational quality • Time-stamped and audited transcriptions to ensure accuracy • Identified and penalized critical failures in model persona delivery

Evaluated multi-turn voice interactions to enhance AI assistant conversational performance. Developed and applied rubric-based criteria to assess adherence to personas, accents, and delivery standards. Conducted transcription auditing and penalized model failures for robust reinforcement learning feedback. • Engaged in natural, multi-turn spoken dialogues • Applied atomic rubrics to assess conversational quality • Time-stamped and audited transcriptions to ensure accuracy • Identified and penalized critical failures in model persona delivery

2026 - Present

Multi-Modal Factuality & AI Safety Evals

Text
Assessed LLM responses for factual correctness and AI safety compliance across complex scenarios. Classified content by claim type and conducted extensive factual verification using expert sources. Provided preference rankings and detailed feedback to improve model tone and regional relevance. • Labeled Objective, Subjective, and Response-based claims in LLM outputs • Evaluated model adherence to safety standards, including high-risk content • Performed in-depth fact-checking and documentation • Maintained a 99% quality benchmark in evaluations

Assessed LLM responses for factual correctness and AI safety compliance across complex scenarios. Classified content by claim type and conducted extensive factual verification using expert sources. Provided preference rankings and detailed feedback to improve model tone and regional relevance. • Labeled Objective, Subjective, and Response-based claims in LLM outputs • Evaluated model adherence to safety standards, including high-risk content • Performed in-depth fact-checking and documentation • Maintained a 99% quality benchmark in evaluations

2026 - Present

Education

L

Leeds City College

Btec Computing, Computiong

Btec Computing
2024 - 2026

Work History

M

Mercor

Digital annotation Expert

Leeds
2025 - 2025