Trigyn has a contractual opportunity for a Lead Specialist - Multilingual Speech and Voice Models. This resource will be working Remotely.
Job Description:
- Speech-to-Speech and Multilingual AI Model Development: Research, develop, fine-tune, and optimize open-source speech and language models to enable natural, real-time voice-to-voice conversational AI capabilities, particularly for under-represented languages and dialects.
- Advanced Model Optimization: Implement advanced model compression techniques, including knowledge distillation, structured pruning, layer dropping, and quantization (PTQ/QAT), to develop computationally efficient models suitable for cloud and resource-constrained edge deployment while preserving model quality.
- Low-Resource Language Adaptation: Develop and apply cross-lingual transfer learning, parameter-efficient fine-tuning (LoRA/PEFT), and other adaptation techniques to improve automatic speech recognition (ASR), speech understanding, and text-to-speech (TTS) capabilities for languages with limited training resources.
- Speech Data Research and Synthetic Data Generation: Research and implement ethical, community-driven methodologies for collecting, annotating, and validating multilingual speech datasets.
- Real-Time Voice AI Optimization: Research and optimize low-latency speech processing, streaming inference, conversational turn-taking, and interruption handling to support natural and responsive voice-based interactions.
- Model Evaluation: Develop evaluation methodologies and benchmarks to measure speech recognition accuracy, speech synthesis quality, conversational performance, language coverage, inference latency, and computational efficiency.
Competencies:
- Professionalism: Demonstrates strong technical expertise and sound judgment in delivering high-quality, reliable solutions.
- Planning and Organizing: Effectively prioritizes tasks, manages time, and meets established deadlines.
- Teamwork: Collaborates efficiently with colleagues across diverse and multidisciplinary teams.
- Communication: Communicates clearly and concisely, including the ability to convey technical information to non-technical stakeholders.
- Accountability: Takes ownership of responsibilities, delivers results in line with agreed objectives, and adheres to organizational standards and policies. System performance, timely issue resolution, and adherence to security standards.
Experience:
- Education: Advanced university degree (Master’s degree or equivalent) in Computer Science, Information Systems, Mathematics, or a related field. A Bachelor’s degree with relevant experience may be accepted in lieu of an advanced degree.
- Minimum of 5 years of experience in research and development of AI-related areas
- Demonstrated experience in leading the development, mentoring, fine-tuning, and evaluating open-source speech, audio, and language models, including ASR, TTS, and speech-to-speech systems.
- Practical knowledge of multilingual speech processing, cross-lingual transfer learning, low-resource language adaptation, and speech dataset development.
- Familiarity with real-time audio streaming, inference optimization, and deploying models in resource-constrained environments
- Experience with knowledge distillation, model quantization, parameter-efficient fine-tuning, and GPU-based model training
- Ability to design experiments, evaluate model performance, document research findings, and collaborate with multidisciplinary teams
- Demonstrated ability to provide technical leadership, guide research and experimentation, review model architectures, and translate research outcomes into practical implementation strategies.
TRIGYN TECHNOLOGIES is a multinational IT services company with resources deployed in 50 countries. TRIGYN is an ISO 9001:2015, ISO 27001:2022 (ISMS) and CMMI Level 5 certified company. TRIGYN has offices in the United States, Canada, Switzerland and India.