Imranullah Wafa

Key Skills

Software

Img Lab

Labelbox

Remotasks

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Video

Top Task Types

Action Recognition

Evaluation Rating

Prompt Response Writing SFT

RLHF

Translation Localization

Freelancer Overview

Architectural-engineering specialist turned AI data trainer with 6 + years of hands-on design, modeling, and technical problem-solving experience. Proven record of creating high-quality training data, rubrics, and evaluations for large language models (LLMs) and reinforcement-learning pipelines. Combines deep domain knowledge in architecture & civil engineering (AutoCAD, Revit, BIM, construction standards) with advanced AI-workflow skills (RLHF, prompt engineering, red-teaming) to improve model reasoning, accuracy, and safety.

Labeling Experience

Multimodal Biscuits – Senior Reviewer (Outlier)

Scale AIImageText GenerationFine Tuning

Worked as a Senior Reviewer and Squad Member on Outlier’s Biscuits with Rubrics project, reviewing over 750 multimodal tasks involving image-text-audio content. Applied detailed rubrics to evaluate model outputs for coherence, reasoning, relevance, and safety. Contributed to rubric development and edge-case testing, ensuring annotation quality through consistent 4★+ reviews. Supported peer onboarding and internal discussions to improve clarity and coverage of difficult task types.

Worked as a Senior Reviewer and Squad Member on Outlier’s Biscuits with Rubrics project, reviewing over 750 multimodal tasks involving image-text-audio content. Applied detailed rubrics to evaluate model outputs for coherence, reasoning, relevance, and safety. Contributed to rubric development and edge-case testing, ensuring annotation quality through consistent 4★+ reviews. Supported peer onboarding and internal discussions to improve clarity and coverage of difficult task types.

2023 - 2024

Cypher RLHF – Response Ranking & Evaluation (Outlier)

Scale AITextClassificationText Generation

Contributed to the Cypher RLHF project by evaluating and ranking over 500 AI-generated responses using Reinforcement Learning from Human Feedback (RLHF) methods. Assessed outputs for instruction following, helpfulness, truthfulness, and relevance. Applied multi-dimensional rubrics and flagged unsafe, biased, or hallucinated content. Collaborated on identifying weak reasoning in multi-turn responses, contributing to continuous improvements in prompt-response alignment. Earned 5★ reviewer ratings for accuracy and speed.

Contributed to the Cypher RLHF project by evaluating and ranking over 500 AI-generated responses using Reinforcement Learning from Human Feedback (RLHF) methods. Assessed outputs for instruction following, helpfulness, truthfulness, and relevance. Applied multi-dimensional rubrics and flagged unsafe, biased, or hallucinated content. Collaborated on identifying weak reasoning in multi-turn responses, contributing to continuous improvements in prompt-response alignment. Earned 5★ reviewer ratings for accuracy and speed.

2022 - 2023

Education

Work History