Machine Learning Engineer - TTS & Expressive Speech

JOB TITLE: Machine Learning Engineer - TTS & Expressive Speech
Company: Ethical Den
Location: Kolkata, West Bengal
Work Mode: On-site only
Employment Type: Full-time
Experience: 2 to 5+ years
Openings: 1
Joining: Immediate to 30 days preferred
ABOUT THE ROLE
Ethical Den is looking for a Machine Learning Engineer specialising in Text-to-Speech, speech synthesis and expressive speech generation.
You will join our AI research and engineering team and work on a confidential speech-technology initiative.
The role combines model research, fine-tuning, dataset development, speech quality and real-time inference performance.
Further project information will be shared with shortlisted candidates subject to confidentiality requirements.
KEY RESPONSIBILITIES
• Research and evaluate modern TTS architectures.
• Fine-tune and adapt pretrained speech models.
• Improve speech naturalness and pronunciation.
• Work on expressive and emotional speech generation.
• Improve prosody, rhythm, pitch, pauses and intonation.
• Work with multi-speaker datasets.
• Develop speaker-conditioning systems.
• Experiment with speaker embeddings.
• Explore lightweight model-adaptation techniques.
• Develop streaming speech-generation pipelines.
• Optimise time to first audio.
• Improve real-time factor.
• Reduce inference latency.
• Optimise GPU memory consumption.
• Design TTS evaluation frameworks.
• Conduct objective and subjective speech evaluation.
• Work closely with speech-data specialists.
• Work with ML systems engineers on inference optimisation.
• Document experiments and model-performance results.
REQUIRED SKILLS
• Python
• PyTorch
• Deep Learning
• Audio ML
• Speech processing
• TTS training and fine-tuning
• Audio preprocessing
• PCM/WAV fundamentals
• Sampling rates
• Spectrograms
• GPU training
• Mixed precision
• Dataset preparation
• Model evaluation
• Linux
• Git
IMPORTANT CONCEPTS
Candidates should understand:
• Prosody
• Phonemes
• Speaker embeddings
• Vocoders
• Streaming synthesis
• Real-time factor
• Time to first audio
• Speaker conditioning
EXPERIENCE WITH ANY OF THE FOLLOWING IS VALUABLE
• XTTS
• VITS
• FastSpeech
• StyleTTS
• Parler-TTS
• Neural codecs
• Diffusion-based speech models
• Flow-based TTS
• LoRA
• ONNX
• TensorRT
EDUCATIONAL QUALIFICATION
Preferred:
B.Tech/B.E./M.Tech/M.Sc. in Computer Science, Electronics, AI/ML, Signal Processing, Speech Technology, Computational Linguistics or a related field.
Research candidates with lower corporate experience may be considered if they can demonstrate meaningful TTS model-development experience.
RESEARCH EXPERIENCE IS WELCOME
We encourage applications from candidates who have:
• Published relevant research
• Released models on Hugging Face
• Contributed to open-source speech projects
• Worked in speech research laboratories
• Independently trained or fine-tuned speech models
API integration with commercial TTS providers alone will not be considered sufficient model-development experience.
APPLICATION DETAILS
Please include:
• Updated CV
• Current location
• Current CTC
• Expected CTC
• Notice period
• GitHub / Hugging Face links
• Publications, if applicable
• Audio samples or model demonstrations, where available
• Short description of relevant speech-model work
CONFIDENTIALITY
This position involves confidential research and product development. Specific product details, datasets and internal architecture will be disclosed only at the appropriate stage of the hiring process and may require an NDA.

Work arrangement
No