Shivika Pandey

Key Skills

Software

AWS SageMaker

Anno-Mage

Appen

Clickworker

CrowdSource

Data Annotation Tech

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Document

Image

Top Label Types

Action Recognition

Bounding Box

Classification

Computer Programming Coding

Prompt Response Writing SFT

Freelancer Overview

I am a Python-specialized AI Trainer and Developer with active experience integrating Generative AI into production SaaS products. Currently working with startups like Vitra Frame and Soul Code, I review Pull Requests (PRs) and optimize AI-driven workflows for video and web tools. With over 50 GitHub repositories and 300+ tracked coding hours in the 2024-25 cycle, I possess the technical expertise to generate high-quality code samples, debug complex Python logic, and evaluate LLM code outputs for syntax, efficiency, and security vulnerabilities. Beyond engineering, I bring a unique "Humanities" edge as a Published Author and National-level Debater. This combination allows me to excel in Reinforcement Learning from Human Feedback (RLHF) tasks that require both technical accuracy and linguistic nuance. I am capable of handling complex data labeling projects—from validating code snippets to synthesizing creative writing prompts—ensuring that AI models are trained on data that is not only factually correct but also culturally and tonally precise.

Labeling Experience

Python Code Generation & RLHF for SaaS AI Agents

Data Annotation TechComputer Code ProgrammingBounding BoxText Generation

Lead the evaluation and refinement of AI-generated Python and JavaScript code for a video editing and website building SaaS (Vitra Frame/Soul Code). Performed expert-level RLHF by reviewing AI-generated Pull Requests (PRs), identifying logic errors, and rewriting inefficient code to create "Gold Standard" training samples. Conducted Red Teaming to identify security vulnerabilities and hallucinations in the model's output. Developed complex prompts to test the AI's ability to handle multi-step reasoning tasks in a production environment.

Lead the evaluation and refinement of AI-generated Python and JavaScript code for a video editing and website building SaaS (Vitra Frame/Soul Code). Performed expert-level RLHF by reviewing AI-generated Pull Requests (PRs), identifying logic errors, and rewriting inefficient code to create "Gold Standard" training samples. Conducted Red Teaming to identify security vulnerabilities and hallucinations in the model's output. Developed complex prompts to test the AI's ability to handle multi-step reasoning tasks in a production environment.

2025

Education

Work History