Fardeen A.

Key Skills

Software

Scale AI

Don't disclose

Label Studio

Top Subject Matter

LLM output evaluation and AI code review for training/feedback loops

Human feedback generation for LLM training (RLHF/RLAIF) and evaluation

NLP dataset annotation and labeling for model training

Top Data Types

Text

Image

Document

Top Task Types

RLHF

Entity (NER) Classification

Freelancer Overview

AI Code Evaluator (Freelance / Self-directed) | Outlier.ai-Aligned Practice. Brings 3+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Scale AI, Don't disclose, and Label Studio. Education includes Bachelor of Technology, Siksha 'O' Anusandhan University (ITER) (2028) and Software Engineering Program, Masai School (2026). AI-training focus includes data types such as Computer Code, Programming, and Text and labeling workflows including Evaluation, Rating, and RLHF.

Labeling Experience

AI evaluation and human-feedback practice (RLHF/RLAIF style)

Don't discloseTextRLHF

Produced instruction-following and preference-labeling style assessments of model outputs, supporting RLHF/RLAIF-aligned training feedback. Identified subtle hallucinations, logical flaws, and factuality issues across math, science, coding, and reasoning prompts. Generated evaluation notes and rubric-driven judgments to strengthen high-quality human feedback for model improvement. • Checked helpfulness, on-topic relevance, and completeness against the user query. • Flagged unsafe or misaligned content via red-teaming and adversarial prompt testing concepts. • Rated and compared multi-response outputs using nuanced quality criteria. • Logged recurring error patterns for downstream dataset and training refinements.

2025 - Present

AI Code Evaluator (Freelance / Self-directed) | Outlier.ai-Aligned Practice

Scale AI

Evaluated hundreds of AI-generated code samples across Python, JavaScript, Java, SQL, and C++ for correctness, efficiency, security, and style. Applied rubric-based comparative preference ranking to judge which model outputs were more accurate, complete, helpful, and instruction-following. Documented hallucinations, factual errors, and alignment failures across coding and reasoning tasks to improve feedback quality. • Designed test suites to expose edge cases and logical errors in LLM-produced code. • Performed security and complexity assessments including vulnerability spotting and code quality checks. • Used structured rubrics to score outputs and support ranking decisions. • Produced error analysis notes for recurring model failure modes and response defects.

2025 - Present

Data Annotation & Labeling (self-directed/NLP labeling practice)

Label StudioTextEntity (NER) Classification

Created or would create structured labeled datasets for NLP training tasks such as classification and NER as part of annotation and evaluation practice. Applied labeling workflows that emphasize data quality review and consistency across samples. Produced labeled outputs intended for conversational AI training and preference-ranking style learning. • Labeled entities for NER-related datasets with quality checks. • Prepared classification labels for NLP instruction and text understanding tasks. • Performed instruction-following and preference labeling style dataset construction. • Conducted data quality review and labeling guideline adherence during dataset creation.

2024 - 2024

Education

Work History