Whisper ASR Fine-Tuning System
Budget: $3,000 – $5,000 USD
I'm looking for an experienced ASR engineer to fine-tune a Whisper-based model for a low-resource language across two milestones:
• Milestone 1: Initial fine-tuning on ~3,000 prepared audio segments (~10 hours, clean and segmented), showing a measurable WER improvement over baseline.
• Milestone 2: Continued training on additional data to reach a target WER of 10% or below, plus documentation and a training setup so a separate full-stack web developer can independently continue improving the model toward 5% WER afterward.
Your scope is strictly the model — a separate full-stack developer handles the platform and integration.
I'm only considering candidates who have already built a similar production ASR system and can provide references. Please describe a specific past project (language, data volume, WER achieved) in your proposal, along with a rough fixed-price estimate for Milestone 1.
• Milestone 1: Initial fine-tuning on ~3,000 prepared audio segments (~10 hours, clean and segmented), showing a measurable WER improvement over baseline.
• Milestone 2: Continued training on additional data to reach a target WER of 10% or below, plus documentation and a training setup so a separate full-stack web developer can independently continue improving the model toward 5% WER afterward.
Your scope is strictly the model — a separate full-stack developer handles the platform and integration.
I'm only considering candidates who have already built a similar production ASR system and can provide references. Please describe a specific past project (language, data volume, WER achieved) in your proposal, along with a rough fixed-price estimate for Milestone 1.