Data Scientist

About Arctan:

At Arctan, were solving a hard problem that matters: Eliminating communication barriers in Global conversations. Our AI product refines speech on the fly, adjusting accents and tonality, helping Customer Facing teams engage better with their customers and build rapport effortlessly across cultures.

Founded in 2024 by IIT Roorkee alumni, Arctan is backed by some of the leading global investors and industry experts. Were a high talent density team, bringing together skillset and expertise across technology, product, and business.

Role Overview:

This is a hands-on, high ownership role. You'll be responsible for the core speech models at Arctan.

As part of the Data Science team, youll have ownership of model direction, experimentation, and deployment strategy.

Its a high-trust role for someone who enjoys both research and execution.

What you'll do:

• Build and train speech-to-speech models
• Use transformer-based and self-supervised architectures
• Work on multilingual, cross-lingual, and style transfer capabilities
• Curate large, diverse datasets
• Evaluate models using WER, accent classification models and latency metrics

You should apply if:

• You have 3+ years of hands-on experience with speech/audio models
• Strong background in ASR, TTS, voice conversion, or related areas
• Deep understanding of architectures like diffusion models, GANs, and attention-based models
• Experience with PyTorch and audio toolkits (ESPnet, torchaudio, Kaldi, etc.)
• Startup mindset: comfort with ambiguity, high agency, fast execution

Nice to Have:

• Experience with multimodal models
• Deployment experience on edge or browser
• Exposure to MLOps practices

Other Details:

• Location: Bengaluru (in-office)
• Type: Full-time