Meghan F.

Key Skills

Software

Internal/Proprietary Tooling

Data Annotation Tech

Google Cloud Vertex AI

SuperAnnotate

Surge AI

Scale AI

LabelImg

Label Studio

AWS SageMaker

Top Subject Matter

LLM code evaluation

Codex

validated multi-language algorithms

Top Data Types

Computer Code Programming

Video

Image

Top Task Types

Bounding Box

Classification

Computer Programming/Coding

Prompt + Response Writing (SFT)

Text Generation

RLHF

Object Detection

Fine-tuning

Evaluation/Rating

Text Summarization

Freelancer Overview

CS grad with 6mo experience in LLM code evaluation & AI alignment. Benchmarked codebase generation, executed RLHF coding, and validated multi-language algorithms. Adept at translating technical requirements into high-quality training datasets for tier-1 AI partner labs

Labeling Experience

AI Software Engineer Fellow (Technical Evaluator)

Computer Code ProgrammingPrompt + Response Writing (SFT)

* Conducted rigorous **code syntax evaluation** and semantic debugging on multi-language code generated by foundational Large Language Models (LLMs). * Utilized **RLHF (Reinforcement Learning from Human Feedback)** frameworks to grade and optimize code correctness, time complexity, and adherence to edge-case specifications. * Designed complex prompts and multi-turn interaction workflows to stress-test model reasoning capabilities in **Python, Java, and C++*

2026 - Present

Generalist AI Trainer

VideoEvaluation/Rating

Create and evaluate high-quality AI training data across technical and general domains, with a focus on coding tasks, reasoning quality, instruction following, and response usefulness. Assess model outputs for clarity, correctness, completeness, and alignment with task requirements. Provide detailed written feedback to improve model behavior and evaluator consistency.

2025 - Present

AI Software Engineer (Technical Evaluator) - Handshake AI

RLHF

Served as a Fellow (Technical Evaluator) conducting syntax and semantic evaluation of multi-language code generated by foundational large language models. Applied RLHF-style grading to improve correctness, performance characteristics, and compliance with edge-case requirements. Built automated verification workflows using testing and prompt-driven stress tests across Python, Java, and C++. • Performed code syntax evaluation and semantic debugging on model-generated repositories • Used RLHF frameworks to assess and optimize output quality • Designed multi-turn prompts to stress-test reasoning and instruction following • Implemented unit testing with PyTest to verify outputs against engineering benchmarks

2025 - Present

AI Software Engineering Fellow (Technical Evaluator) — Handshake AI

RLHF

Evaluated and rated multi-turn, model-generated code outputs using RLHF methodologies to improve correctness, performance, and edge-case compliance. Performed semantic debugging and security/vulnerability checks to ensure responses met strict engineering benchmarks. Produced quality-controlled judgments and test-based validations suitable for partner-lab dataset improvement. • Graded code correctness and adherence to edge-case specifications • Designed prompts and multi-turn workflows in Python, Java, and C++ • Authored PyTest unit tests to verify outputs against engineering benchmarks • Documented logical fallacies, security vulnerabilities, and syntax discrepancies for iteration

2025 - Present

Pattern Extraction

OtherImageObject DetectionAction Recognition

This task involves creating AI training data by extracting patterns from images of objects such as clothing, furniture, or wallpaper and transforming them into new styles. The process includes cropping the target object, isolating and extracting its pattern as a flat, tileable texture, and applying a new pattern while keeping the object’s shape and background intact. Additionally, a reversal prompt is written to allow the AI model to reapply the original pattern onto the transformed object, ensuring the training data demonstrates both pattern extraction and reapplication. Attention to detail, color accuracy, style fidelity, and clear prompt writing are essential to producing high-quality training examples for generative image models.

2025

Show more

Education

Work History