Lead Research Scientist Voice AI

About the role

ToumAI develops voice AI for underserved languages and dialects, focusing on open‑source, data‑centric solutions. The team tackles high‑quality voice AI under linguistic complexity and data scarcity, such as code‑switching between Moroccan Darija and French or noisy low‑resource Swahili calls.

Key responsibilities

• Build and adapt text‑to‑speech (TTS) models emphasizing naturalness, pronunciation and prosody across dialects.
• Improve multilingual, code‑switch‑aware speech‑to‑text (STT) for real‑world conversational speech.
• Design transfer‑learning and fine‑tuning strategies that accelerate new language development.
• Select and benchmark open‑source foundations for specific languages and use cases.
• Develop data‑centric strategies for scarce or fragmented training data.
• Define evaluation methodologies that combine automated metrics with native‑speaker assessments.

Required profile

• PhD in machine learning, NLP, speech processing, computational linguistics or related field, with 2–5 years of research or applied experience (exceptional recent graduates are welcome).
• Deep understanding of modern ML, experimental design and a data‑centric approach to model improvement.
• Proven ability to train, fine‑tune and evaluate deep‑learning models at scale.
• Experience translating research into production‑ready systems and communicating decisions to engineers and stakeholders.

Required skills

• Python programming.
• Modern deep‑learning frameworks (e.g., PyTorch, TensorFlow).
• Speech synthesis (TTS) and automatic speech recognition (ASR) techniques.
• Transfer learning, fine‑tuning and representation learning.
• Large language models (LLMs) integration within speech pipelines.