Career

Key Responsibilities

• Research, design, and develop state-of-the-art Text-to-Speech (TTS) systems for natural, real-time human-computer interaction.

• Train, fine-tune, and optimize neural TTS models (e.g., Tacotron, FastSpeech, VITS, WaveNet) to achieve high-quality, expressive, and low-latency speech synthesis.

• Build multilingual and multi-speaker TTS pipelines, with controllable style, prosody, and emotional tone.

• Collaborate with ASR and NLP/LLM teams to ensure smooth end-to-end voice interaction systems.

• Explore and integrate multimodal approaches where TTS works with text, speech, and vision for richer interaction.

• Develop evaluation frameworks and metrics for TTS quality (MOS, intelligibility, latency).

• Research and implement model compression, quantization, and deployment techniques for real-time inference on edge devices.

• Stay updated on the latest research in speech synthesis, voice cloning, and generative models (e.g., diffusion-based speech synthesis).
  
  

Requirements

• Bachelor’s, Master’s, or PhD in Computer Science, AI, Speech Processing, Computational Linguistics, or related field.

• Strong programming skills in Python, with experience in PyTorch/TensorFlow.

• Solid knowledge of deep learning, signal processing, and speech synthesis.

• Hands-on experience with TTS frameworks (e.g., ESPnet, NVIDIA NeMo, Fairseq, Coqui TTS).

• Understanding of vocoder models (WaveNet, HiFi-GAN, WaveGlow, etc.).

• Familiarity with ASR, NLP, or LLMs to build integrated voice-AI solutions.

• Knowledge of deployment optimization: low-latency inference, edge AI, GPU/TPU acceleration.
  
  

Preferred Qualifications

• Experience in deploying real-time TTS systems in production.

• Background in multilingual/multispeaker TTS and voice cloning.

• Research or practical experience with expressive/emotional TTS.

• Knowledge of reinforcement learning from human feedback (RLHF) for speech quality improvement.

• Expertise in distributed training and large-scale model serving.