Cupidwritings

Key Skills

Software

Don't disclose

Top Subject Matter

E-Commerce & Product Technology

Generative AI and Machine Learning

Natural Language Processing & Linguistics Application

Top Data Types

Computer Code Programming

Text

Image

Top Task Types

RLHF

Evaluation/Rating

Freelancer Overview

Through building Lensfuse, an AI-powered product image generation tool, and a separate text-to-image web application, I gained direct, practical experience working within generative AI pipelines as both a developer and an evaluator. On Lensfuse, I designed the full backend architecture using PHP and Laravel, integrated third-party AI image generation APIs, and managed the complete input-to-output workflow, giving me an intimate understanding of how data structure, prompt quality, and input framing directly shape what an AI model produces. A significant part of that work involved writing and iteratively refining text prompts to guide the model toward commercially viable, studio-quality results, then assessing each output for prompt-output alignment, visual accuracy, and consistency across different product types. This cycle of crafting inputs, rating outputs, and correcting failure patterns is the same core loop that powers RLHF and evaluation-based labeling workflows. Beyond the technical side, I hold formal academic training in linguistics, including coursework in syntax, X-bar theory, semantics, and language structure, which gives me a precise, analytical lens when working with natural language data. I understand how sentence structure, word choice, and semantic specificity affect how language is interpreted, whether by a human or a model. Combined with my background as a published writer, I bring strong written communication skills and an eye for nuance that is especially valuable in text classification, annotation quality review, and natural language tasks. I am technically fluent in structured data formats, version-controlled workflows via Git and GitHub, and maintaining output consistency at scale, making me well-equipped to contribute meaningfully to any AI training data project from day one.

Labeling Experience

AI Text Prompt Engineering & Quality Evaluation for Generative Image Models

TextRLHF

Developed and evaluated AI-generated outputs for Lensfuse, a generative AI tool designed to transform raw product photos into studio-quality images for e-commerce use cases. The core work involved iterative prompt engineering such as crafting, testing, and refining text inputs to guide image generation models toward consistent, commercially usable outputs. Tasks included evaluating output quality against real-world standards (lighting accuracy, background coherence, product fidelity), identifying failure patterns in generated results, and adjusting prompt structure to correct them plus, a process closely aligned with RLHF-style human feedback loops. Quality measures adhered to included visual accuracy, prompt-output alignment, and consistency across varied product types. Concurrently built a text-to-image application, deepening practical knowledge of how natural language inputs are interpreted by generative models and how phrasing, specificity, and semantic clarity affect output quality at scale.

2025 - 2026

Education

Work History