Rosseliet Galarraga
Audio AI Specialist | Speech Data, Quality & Annotation
Contratar a esta persona
Envía una oferta de trabajo directamente a este candidato
Sobre mí
Audio AI and data annotation specialist with hands-on experience supporting speech and voice datasets for AI model training. Experienced in audio quality review, transcription QA, timestamp validation, VAD/segmentation, speaker diarization, pronunciation checks, metadata review, and evaluation of real and synthetic audio. Strong ability to identify subtle audio defects and annotation inconsistencies, apply detailed quality guidelines, document edge cases, and deliver precise, actionable feedback.
Comfortable working with structured data in spreadsheets, CSV and JSON, as well as audio inspection and QA tools.
Experiencia
Audio AI & Data Annotation Specialist at Pixzelle Studio / Pixzelle AI (2023-11 – 2026-01)
- Supported development and quality assurance of speech datasets used for AI model training across TTS, transcription, ASR, speech-to-speech and conversational voice workflows.
- Reviewed audio against technical, linguistic and task-specific standards; made approval, rejection and revision decisions based on established quality guidelines.
- Detected background noise, clipping, distortion, plosives, echo, low signal, segmentation errors, transcript mismatches and speaker-label inconsistencies.
- Performed speech annotation and QA including transcription review, timestamp validation, VAD/segmentation, speaker diarization, pronunciation checks and metadata validation.
- Evaluated naturalness, intelligibility and consistency across real and synthetic speech, including accent and dialect variation in English and Spanish.
- Recorded specification-compliant speech from a controlled environment and reviewed recordings for clarity, consistency and dataset usability.
- Documented edge cases, refined review rubrics and SOPs, and provided precise written feedback to improve annotation consistency and downstream model-training data quality.
- Worked with structured datasets and metadata using spreadsheets, CSV and JSON; used Audacity and Praat for audio inspection and basic Python, Bash and SQL for QA checks and dataset review.
- Collaborated with technical, research and operations stakeholders to translate model requirements into practical data specifications and evaluation criteria while maintaining compliant handling of voice data.
Content Reviewer / Quality Assurance
- Reviewed high-volume content against detailed policies and quality standards, identified inconsistencies and edge cases, documented decisions, and maintained accuracy and consistency across review workflows.
Educación
High School / Technical Studies in Information Technology – Raúl Cepero Bonilla Technical Institute (2022)
