AI Engineer, Speech-to-Text

givaaerotech; India; Remote; INR 1,800,000 - 3,200,000; Part time

Job summary

givaaerotech is developing GiVa™ HI Smart Glasses to provide real-time captioning for hearing-impaired users. This role focuses on evaluating, prototyping, and optimizing speech-to-text systems for real-time captioning in a remote-friendly setup.

The ideal candidate has 4–5 years in speech recognition or audio AI, familiarity with Whisper and wav2vec, and an interest in accessibility tech. You will work with a cross-functional team to define integration requirements and select the best STT

Qualifications

• 4–5 years in speech recognition, NLP or audio AI.
• Familiarity with Whisper, wav2vec, Vosk, DeepSpeech, Hugging Face or cloud STT APIs.
• Understanding of latency, noise reduction and audio preprocessing.

Responsibilities

• Evaluate cloud-based and edge-based speech-to-text models.
• Compare STT solutions for accuracy, latency, cost and language support.
• Test performance in noisy environments and real-world speech conditions.
• Build an audio preprocessing and transcription pipeline.
• Explore Indian language and multilingual captioning feasibility.
• Collaborate with embedded team to define integration requirements.
• Recommend the best STT architecture for prototype and MVP.

Skills

Speech recognition

NLP

Latency optimization

Prototype pipeline design

Tools

Whisper

wav2vec

Vosk

DeepSpeech

Hugging Face

Job description

Real-time captioning for GiVa™ HI Smart Glasses

We are building GiVa™ HI Smart Glasses, an assistive wearable for real-time speech-to-text support for hearing-impaired users. This role evaluates, prototypes and optimises speech-to-text systems for real-time captioning.

Team AI

Experience 4-5 years

Engagement Consultant or contract initially

Location Remote possible

Stage STT model evaluation and prototype pipeline

What you will work on

• Evaluate cloud-based and edge-based speech-to-text models.
• Compare STT solutions for accuracy, latency, cost and language support.
• Test performance in noisy environments and real-world speech conditions.
• Build an audio preprocessing and transcription pipeline.
• Explore Indian language and multilingual captioning feasibility.
• Work with the embedded team to define integration requirements.
• Recommend the best STT architecture for prototype and MVP.

Ideal profile

• 4-5 years in speech recognition, NLP, audio AI or voice systems.
• Familiarity with Whisper, wav2vec, Vosk, DeepSpeech, Hugging Face or cloud STT APIs.
• Understanding of latency, noise reduction, audio preprocessing and model evaluation.
• Interest in accessibility technology is a strong plus.

Preferred tools

Python, PyTorch, TensorFlow, Hugging Face, Whisper, Google Speech-to-Text, Azure Speech and AWS Transcribe.