Elias M

Audio QA and data quality specialist with hands-on experience reviewing, annotating, and validating speech datasets for AI/ML training pipelines, combined with a software engineering background in building evaluation and data-validation tooling. Direct experience as a QA Audio Reviewer and Transcriber on a production speech-data project, screening recordings for noise, clipping, distortion, segmentation errors, and transcript/annotation consistency. Strong auditory judgment paired with the ability to build lightweight Python and SQL tooling for QA automation, anomaly detection, and reporting.

Comfortable working across structured data formats (CSV, JSON) and task-tracking systems, and translating loosely defined data-quality requirements into consistent, repeatable review standards.

Speech Data Annotation Specialist at Alignturn (2026-01 – 2026-06)

Served as QA Audio Reviewer (Jan–Mar 2026), reviewing recorded speech against technical and linguistic quality standards and making approve/reject/revise decisions to ensure audio met specification for AI/ML training pipelines. Transitioned to Transcriber (Apr–Jun 2026), producing accurate, specification-compliant transcriptions of spoken audio to support speech-to-text dataset development.

• Screened audio for common quality defects — background noise, clipping, distortion, plosives, echo, and low signal levels — and flagged files requiring rework
• Checked segmentation, timestamps, and transcript/speaker-label alignment for consistency across paired audio-annotation deliverables
• Maintained consistent accuracy and formatting standards across a high volume of audio files under production deadlines

AI Evaluation Specialist / LLM QA Engineer at Independent (2022-01 – Present)

Designed and implemented evaluation pipelines and review rubrics to assess reasoning quality, instruction-following, and robustness across model outputs — directly analogous to defining and applying quality standards for a dataset.

• Built Python tooling for automated scoring, structured evaluation, and comparative quality analysis, including anomaly detection for identifying inconsistent or low-quality samples
• Documented systematic failure modes and edge cases, feeding findings back into iterative improvement of evaluation guidelines and SOPs
• Performed data validation, noise filtering, and QA checks on AI training datasets to ensure integrity across data, labels, and metadata

Software Developer / ML Systems Contributor at Freelance

• Built Python utilities for dataset validation, preprocessing, and experiment automation across ML pipelines
• Developed batch-evaluation tooling for structured comparison of outputs across systems, supporting consistent, repeatable QA review
• Debugged data inconsistencies affecting downstream model training quality

Physics / Computational Science