Fahad A.

Key Skills

Software

Mercor

Google Cloud Vertex AI

OneForma

OpenCV AI Kit (OAK)

Redbrick AI

Remotasks

Scale AI

Snorkel AI

SuperAnnotate

Surge AI

Don't disclose

Top Subject Matter

Frontier LLM/VLM evaluation and technical AI safety

VLM behavior evaluation and AI safety interventions

Legal Services & Contract Review

Top Data Types

Text

Document

Computer Code Programming

Top Task Types

Fine-tuning

RLHF

Bounding Box

Polygon

Segmentation

Classification

Polyline

Cuboid

Object Detection

Text Generation

Question Answering

Text Summarization

Red Teaming

Evaluation/Rating

Computer Programming/Coding

Data Collection

Transcription

Function Calling

Point/Key Point

Prompt + Response Writing (SFT)

Freelancer Overview

Mercor | Data Scientist - AI/ML Expert, Contract | Remote, USA / Scottsdale, AZ. Brings 5+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Mercor and Don't disclose. Education includes Master of Science, Arizona State University (2026) and Bachelor of Science, Arizona State University (2022). AI-training focus includes data types such as Computer Code, Programming, and Text and labeling workflows including Evaluation and Rating.

Labeling Experience

Mercor | Data Scientist - AI/ML Expert, Contract | Remote, USA / Scottsdale, AZ

Mercor

Designed frontier model evaluation workflows using rubric-based grading and reference solutions to systematically expose hallucinations, brittle code, and missed edge cases. Created benchmark-quality prompts and correction rationales to enable reproducible technical scoring across diverse test conditions. Led stress testing of AI outputs by iterating through data cleaning, feature engineering, experiment design, and production-style debugging workflows. • Built grading rubrics and reference solutions for evaluation • Developed evaluation prompts and golden sets • Performed failure analysis for edge cases and reasoning gaps • Used Python/SQL/ML tasks to probe model assumptions

2026 - Present

Arizona State University | Graduate Researcher, Interpretability & AI Safety

Don't discloseText

Built behavior evaluation pipelines for VLM safety research, including discovery splits, causal prescreening, steering tests, and transfer benchmarks. Performed safety-focused negative-result checks to detect sparse trap-spotting behavior and characterize model responses under intervention. Quantified intervention effectiveness using AUROC-based feature detection and smart-rate improvements without retraining. • Created evaluation splits and transfer benchmarks • Conducted causal prescreening and steering tests • Identified trap-spotting features (AUROC > 0.99) • Measured Qwen2.5 VLM smart-rate gains from interventions

2025 - 2026

Education

Work History