Robert Melrose

AI Speech Quality Expert | Audio Evaluation | AI Model Trainer

Profile

Full stack software engineer, AI model trainer, audio producer, and voice performer with 10+ years of engineering experience and recent hands-on work in AI evaluation, audio-aware RAG pipelines, speech transcription, audio segmentation, model review, and speech-to-structure product development. Strong fit for speech aesthetics work requiring discriminative listening, annotation QA, guideline development, and reliable evaluation signal for SFT, RLHF, autoeval, and reward-model workflows.

Experience

Founder and Lead Engineer

Oct 2025 - Present
Devs Miami LLC • Miami, FL

Designed and deployed production AI web applications and speech-to-structure workflows.

  • Designed and deployed WingIt, a production full-stack AI web application, owning planning, testing, CI/CD, container delivery, observability, analytics, funnels, dashboards, and session replay.
  • Building speech-to-diagram and speech-to-presentation workflows where live conversation must be interpreted into structured visual output while preserving speaker intent, context, task adherence, and audience readability.
  • Write product specs, behavior guidelines, QA criteria, and acceptance tests for AI-generated diagrams, slide content, multimodal presentation workflows, and model-assisted visual communication.

Sr. Software Engineer and AI Model Trainer

Jun 2025 - Present
Snorkel AI • Remote

Python-first contractor building audio-aware AI evaluation and RAG workflows.

  • Built evaluation harnesses, RAG-style workflows, and audio-aware pipelines using Whisper, Librosa, FastAPI, Pydantic, SQLAlchemy, embeddings, structured logging, and graph databases.
  • Built STT transcription, audio segmentation, and feature-extraction services to generate high-signal data for evaluating and fine-tuning code-capable language models.
  • Refined agentic planning, tool selection, reusable task execution, and structured output patterns for AI evaluation systems.

Sr. Software Engineer and AI Model Trainer

Nov 2025 - Present
Sepal AI • Remote

AI model training contractor and occupational reviewer.

  • Selected for an accelerated experts team and promoted to occupational reviewer for software development and AI model training.
  • Review model outputs and contributor work for correctness, rubric adherence, clarity, consistency, and useful feedback while guiding junior engineers through code reviews and technical standards.

Full Stack Web Developer and AI Model Trainer

Dec 2023 - Jun 2025
Scale.ai • San Francisco, CA / Remote

Benchmarked top LLMs and reviewed AI model training data for software development tasks.

  • Fine-tuned AI-assisted code generation, bug detection, debugging, and solution explanation workflows on foundational models.
  • Applied reinforcement-learning evaluation patterns to debugging, performance, and parallelism tasks in Python and C++; promoted to senior reviewer and helped onboard junior developers.

Software Engineer I

Jun 2012 - Nov 2014
Seagate Technology • Longmont, CO

Software engineer in reliability with acoustical test tooling experience.

  • Developed Windows Forms applications used for acoustical testing of hard disk drives and maintained Python, C++, and Java reliability tooling.
  • Wrote automated reliability tests and established coding/documentation standards across engineering teams.

Projects

AI Audio Analysis RAG Prototype

RAG system for audio analysis using Whisper transcription, Librosa segmentation, multimodal LLMs, knowledge graphs, and MCP tool usage.

  • Focuses on how audio is segmented, interpreted, retrieved, summarized, and judged by model-assisted workflows.

Web-based audio mastering tool for producers to A/B test how music will sound across streaming normalization targets.

  • Demonstrates listening judgment, loudness perception, playback normalization, producer workflow empathy, and practical understanding of perceived audio fidelity.

General STEM expert contributor and senior reviewer.

  • Reinforced rubric-based review, consistency, and high-quality expert evaluation habits.

Education

B.S. in Computer Science

Colorado State University

A.A. in Photography

Colorado Mountain College

Certifications

CISSP

ISC2

CEH v12

EC-Council

CompTIA Security+

CompTIA

AZ-104 Azure Administrator

Microsoft

AZ-900 Azure Fundamentals

Microsoft

Skills

Speech and Audio

critical listening • speech aesthetics • vocal tone • inflection • pacing • sibilance • noise floor • transients • compression • fidelity • scripted vs spontaneous speech • emotional nuance • conversational intent • Ableton Live • audio production • voice performance • Room EQ Wizard • EVE Audio SC203 • Audio Technica ATH-M50x • Behringer ECM8000

AI Evaluation

SFT • RLHF • model output evaluation • annotation QA • rubric design • senior review • reward-model signal • autoeval support • RAG workflows • Whisper • Librosa • embeddings • graph databases • agentic tool-calling • calibration feedback

Engineering

Python • TypeScript • JavaScript • C# • C++ • FastAPI • Pydantic • SQLAlchemy • Next.js • React • Tailwind CSS • Docker • GitHub Actions • OpenAPI • REST APIs • observability • ASP.NET Core • OAuth/OIDC