Data Scientist (VoiceAI)

About the Role:

We are seeking a highly experienced Voice AI /ML Engineer to lead the design and deployment of real-time voice intelligence systems. This role focuses on ASR, TTS, speaker diarization, wake word detection, and building production-grade modular audio processing pipelines to power next-generation contact center solutions, intelligent voice agents, and telecom-grade audio systems.

You will work at the intersection of deep learning, streaming infrastructure, and speech/NLP technology, creating scalable, low-latency systems across diverse audio formats and real-world applications.

Key Responsibilities:

Voice & Audio Intelligence:

• Build, fine-tune, and deploy ASR models (e.g., Whisper, wav2vec2.0, Conformer) for real-time transcription.
• Develop and finetune high-quality TTS systems using VITS, Tacotron, FastSpeech for lifelike voice generation and cloning.
• Implement speaker diarization for segmenting and identifying speakers in multi-party conversations using embeddings (x-vectors/d-vectors) and clustering (AHC, VBx, spectral clustering).
• Design robust wake word detection models with ultra-low latency and high accuracy in noisy conditions.

Real-Time Audio Streaming & Voice Agent Infrastructure:

• Architect bi-directional real-time audio streaming pipelines using WebSocket, gRPC, Twilio Media Streams, or WebRTC.
• Integrate voice AI models into live voice agent solutions, IVR automation, and AI contact center platforms.
• Optimize for latency, concurrency, and continuous audio streaming with context buffering and voice activity detection (VAD).
• Build scalable microservices to process, decode, encode, and stream audio across common codecs (e.g., PCM, Opus, -law, AAC, MP3) and containers (e.g., WAV, MP4).

Deep Learning & NLP Architecture:

• Utilize transformers, encoder-decoder models, GANs, VAEs, and diffusion models, for speech and language tasks.
• Implement end-to-end pipelines including text normalization, G2P mapping, NLP intent extraction, and emotion/prosody control.
• Fine-tune pre-trained language models for integration with voice-based user interfaces.

Modular System Development:

• Build reusable, plug-and-play modules for ASR, TTS, diarization, codecs, streaming inference, and data augmentation.
• Design APIs and interfaces for orchestrating voice tasks across multi-stage pipelines with format conversions and buffering.
• Develop performance benchmarks and optimize for CPU/GPU, memory footprint, and real-time constraints.

Engineering & Deployment:

• Writing robust, modular, and efficient Python code
• Experience with Docker, Kubernetes, cloud deployment (AWS, Azure, GCP)
• Optimize models for real-time inference using ONNX, TorchScript, and CUDA, including quantization, context-aware inference, model caching.
• On device voice model deployment.

Why join us?

• Impactful Work:

  Play a pivotal role in safeguarding Tanla's assets, data, and reputation in the industry.

• Tremendous Growth Opportunities:

  Be part of a rapidly growing company in the telecom and CPaaS space, with opportunities for professional development.

• Innovative Environment:

  Work alongside a world-class team in a challenging and fun environment, where innovation is celebrated.

www.tanla.com