ML Validation & Fine-Tuning Specialist
Summary: Drive the iterative improvement of the core BirdNET model by managing the data lifecycle, conducting targeted fine-tuning, and performing advanced model validation. This role is responsible for maximizing model performance by incorporating new labels, ensuring high-quality data splits, and settling on the most performant model version for the project.
Duties & Responsibilities
- Lead the collection, labeling, and preprocessing of new acoustic data, including local Costa Rican bird species and critical human activity sounds.
- Design and execute targeted fine-tuning experiments on the existing BirdNET model architecture, analyzing how different learning rates and data mixes impact prediction accuracy and specificity.
- Develop and implement advanced evaluation metrics and testing methodologies to rigorously assess model performance across different data distributions and classes.
- Collaborate with the Data Pipeline Team (Team 3) to integrate new data processing steps and ensure data quality standards are met.
- Produce clear reports detailing model lineage, evaluation results, and the rationale for selecting the final model version for optimization.
Skills & Qualifications (Required)
- Python and standard data science libraries (Pandas, NumPy).
- Proven experience in data processing, cleaning, and labeling for ML tasks.
- Experience with model fine-tuning and transfer learning.
- Strong focus on data integrity and quality assurance throughout the ML lifecycle.
- Methodical and detail-oriented approach to data management and experimental validation.
Skills & Qualifications (Preferred)
- Experience with specialized audio data formats and tools.
- Practical knowledge of advanced ML evaluation metrics beyond simple accuracy.
- Experience specifically with transfer learning in audio or vision models (e.g., using pre-trained weights).
- Strong communication skills for coordinating data needs with other teams