Vishal K.

Key Skills

Software

Appen

Labelbox

Mindrift

OneForma

Scale AI

Telus

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Document

Text

Top Task Types

Computer Programming/Coding

Evaluation/Rating

Prompt + Response Writing (SFT)

Question Answering

RLHF

Freelancer Overview

I specialize in high-precision AI training data and model evaluation, with hands-on experience at TELUS International, Scale AI, and Aligner. My core work involves designing adversarial, PhD-level problem sets in mathematics and computer science to stress-test state-of-the-art models, then evaluating, correcting, and adjudicating model outputs with clear, research-grade rationales. At Scale AI, I contributed to the Human Frontier Collective project, executing multi-stage labeling—prompt design, gold-standard authoring, rubric creation, and quality control—to raise dataset difficulty and signal quality. My strengths include adversarial dataset design, rigorous error taxonomy and rubrics, multi-annotator consistency, and fast iteration from guidelines to production-ready data. I’m currently pursuing a PhD in Computer Science & Engineering at IIT Roorkee, India which gives me deep domain expertise for crafting unambiguous ground truth in mathematically intensive tasks. This blend of research depth and meticulous annotation elevates model robustness and evaluation fidelity.

Labeling Experience

Human Frontier Collective

Scale AITextQuestion Answering

The Human Frontier Collective (HFC) is Scale AI’s part-time, research-oriented program that brings together PhDs, postdocs, and advanced domain experts to directly shape frontier AI. Members design challenging, domain-specific problem sets to “stress-test” cutting-edge generative models, provide expert evaluations and critiques, collaborate with Scale researchers, and may co-author research outputs (reports, talks, webinars). Role listings describe responsibilities like creating and critiquing complex problem sets, analyzing model behavior, informing research directions, and participating in interdisciplinary sessions.

2025

Reasoning Annotations

TelusTextQuestion Answering

Contributed to TELUS International’s Reasoning Annotation (Math) program—designing multi-skill math prompts and performing step-wise error annotation/correction on LLM solutions to create high-fidelity gold data for model training and evaluation (remote, contractor role).

2025 - 2025

Education

Work History