AI TRANSCRIPTION & WER SPECIALIST

About The Opportunity A fast-scaling AI-first startup in the speech technology and conversational AI sector, we build next-gen transcription engines that power voice assistants, call centers, and accessibility tools across global markets. Our team focuses on reducing Word Error Rate (WER) through data science, acoustic modeling, and linguistic refinement — enabling near-human accuracy for multilingual, noisy, and domain-specific speech.

Role & Responsibilities

• Optimize ASR pipelines to reduce WER across noisy, accented, and domain-specific audio (call center, medical, legal).
• Analyze transcription errors and recommend targeted data, feature, or model improvements to boost accuracy.
• Curate, label, and augment speech datasets — balancing coverage, diversity, and domain relevance.
• Run AB tests on model variants, measure WER impact, and document performance trade-offs.
• Collaborate with ML engineers to integrate WER feedback loops into training and inference workflows.
• Define and automate WER evaluation benchmarks for continuous model improvement.

Skills & Qualifications

• Must-Have

• WER analysis
• ASR systems
• Speech dataset curation
• Python
• Speech preprocessing
• Acoustic modeling basics
• AB testing

• Kaldi or Whisper

• Preferred

• Experience with Hugging Face Transformers
• Kaldi pipelines
• Punctuation restoration

Benefits & Culture Highlights

• 100% remote with flexible hours — work from anywhere in India
• Impact-driven culture — your optimizations directly improve real-world ASR performance
• Access to the latest speech AI tools and research — stay at the bleeding edge

Skills: accessibility,data,python,pipelines,access,modeling,speech,acoustic modeling,transcription,acoustic