Ayush S.

Key Skills

Software

Scale AI

Other

Mercor

CrowdSource

Labelbox

Appen

Top Subject Matter

Generative AI for coding language and mathematical reasoning

Industrial cybersecurity (OT/PLC deception environments

Data Analyst/ Evaluator

Top Data Types

Text

Document

Computer Code Programming

Top Task Types

RLHF

Data Collection

Classification

Computer Programming/Coding

Evaluation/Rating

Freelancer Overview

AI Trainer (Hindi & Mathematics) - Outlier (Scale AI). Brings 4+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Scale AI and Other. Education includes Master's Degree in Intelligent Systems, Universitat Jaume I (UJI) (2026) and Postgraduate Diploma in Applied Statistics, Indira Gandhi National Open University (IGNOU) (2024). AI-training focus includes data types such as Text, Computer Code, and Programming and labeling workflows including RLHF and Data Collection.

Labeling Experience

Evaluator

Computer Code ProgrammingEvaluation/Rating

Worked as a freelance Coding Evaluator with Mercor, reviewing and assessing AI-generated coding responses for accuracy, logic, clarity, and correctness. Evaluated solutions in programming-related tasks, identified errors in code, checked whether outputs followed instructions, and provided structured feedback to improve model performance. Applied knowledge of Python, SQL, data analysis, debugging, and problem-solving to judge code quality, reasoning, and final answers. This role strengthened my skills in AI evaluation, technical review, prompt understanding, and quality assurance for machine learning systems.

2026 - Present

MSc Researcher - Cybersecurity (Master's Thesis) — Cuatroochenta S.A.

OtherData Collection

Design and build an industrial OT “Digital Shadow” module that captures real-time PLC telemetry and replicates it inside deception (honeynet) environments. Profile real PLC behavior over Modbus and Profinet protocols and reproduce the behavior on virtual assets so decoy telemetry matches production. Implement a Python-based duplication engine with stages for profiling, replication, deployment, and live-network validation. • Capture and synchronize protocol-aware industrial telemetry for training/validation purposes. • Create realistic decoy behavior by profiling time-series telemetry patterns. • Validate duplication/replication outputs against live OT network traffic. • Work in a SOC-led environment with supervision from SOC leadership.

2026 - Present

AI Trainer (Hindi & Mathematics) — Outlier (Scale AI)

Scale AITextRLHF

Generate, evaluate, and rank Hindi and mathematical reasoning model outputs to improve large language models. Apply domain expertise to assess factuality, reasoning correctness, and relevance of generated content. Contribute to RLHF-style data curation workflows for generative AI system improvement. • Perform quality evaluation and comparative ranking of model responses. • Use language/domain knowledge (Hindi and mathematics) to judge correctness and relevance. • Support reinforcement-learning-from-human-feedback style dataset curation. • Iterate on training data based on evaluation outcomes.

2024 - Present

Education

Work History