Senior Audio AI Engineer – TTS​/Speech Synthesis

At Awarri, our mission is to enable the development and adoption of frontier technology across Africa, starting in Nigeria. We are building inclusive AI technologies—from LLMs to speech models—that reflect and empower African languages and cultural contexts.

Why Join Awarri?

• Be part of a pioneering initiative shaping the future of AI in Africa.
• Work on impactful projects that center real-world representation and inclusivity.
• Collaborate with a passionate, globally distributed team of engineers, linguists, and researchers.

As a Senior Audio AI Engineer at Awarri, you will play a pivotal role in advancing the naturalness and quality of our Text-to-Speech (TTS) systems, focused on African languages and accents. We’re seeking an engineer who understands the intricacies of prosody, rhythm, and speech alignment—and is excited to push the boundaries of audio AI in a meaningful cultural context.

This role is best suited for a specialist with deep experience in speech technologies and a passion for building expressive, production-ready TTS models. You’ll be joining a collaborative, mission-driven team dedicated to shaping the future of generative audio systems in Africa.

• Optimize neural TTS models for prosody, pacing, and expressiveness (e.g., Tacotron 2, Fast Speech 2, Glow-TTS, VITS).
• Improve duration prediction and phoneme-to-frame alignment using forced aligners or prosody-aware training.
• Incorporate punctuation and linguistic markers into the model pipeline to improve natural flow.
• Implement and fine-tune transformer-based architectures for speech synthesis and text-to-speech tasks.

• Evaluate and fine-tune neural vocoders (e.g., HiFi-GAN, Wave Glow) to match desired voice characteristics and audio quality.
• Identify and correct audio artifacts or inconsistencies in generated speech.
• Optimize speech processing pipelines for efficiency and real-time performance.

• Lead both objective (e.g., duration errors, pitch contours) and subjective (e.g., MOS scoring) evaluations of TTS quality.
• Collaborate with linguistic teams to benchmark pronunciation accuracy in Nigerian languages.
• Develop automated testing frameworks to validate speech synthesis quality at scale.

• Prepare the TTS system for product integration by improving inference speed and robustness.
• Support the deployment of models across various platforms (cloud, mobile, embedded).
• Optimize model inference using VLLM for efficient deployment.
• Build APIs and backend services for TTS deployment using FastAPI and Flask
  .
• Implement and manage data pipelines and storage solutions using MongoDB and MySQL
  .

• Proficiency in Python and Type Script for model development and backend integration.
• Experience with transformer-based models for speech synthesis and NLP.
• Strong background in machine learning frameworks such as Tensor Flow or PyTorch.
• Experience in designing scalable AI-driven applications.
• Familiarity with FastAPI
  , Flask
  , and cloud-based deployment environments.
• Knowledge of database management using MongoDB and MySQL
  .

• 3+ years of experience developing and deploying TTS or speech generation systems (bonus for low-resource languages).
• Deep knowledge of at least one neural TTS architecture and related vocoders.
• Proficiency with PyTorch, Tensor Flow, or JAX for building and training models.
• Experience with audio processing tools (e.g., librosa, Praat, torch audio).

• Experience working with multilingual or low-resource speech data.
• Familiarity with phonetics/phonology, especially as it relates to prosody and rhythm.

• Experience building scalable training and evaluation pipelines.
• Ability to debug complex model behavior and iterate quickly toward product quality.
• Comfort working remotely and asynchronously with interdisciplinary teams.

• Prior work on African language speech systems or expressive TTS in non-English languages.
• Interest in linguistic or cultural technology in the African context.
• Contributions to open-source TTS or audio AI tools.
• Experience with emotion modeling or speaker adaptation.