Lead Specialist, AI Scientist

Role Description

The AI Scientist / Engineer - Speech Language Models leads the design, development, and evaluation of AI systems for language assessment, real-time feedback, and skills evaluation. The role focuses on speech recognition, spoken language modeling, and automated scoring, ensuring models are accurate, reliable, fair, and scalable across learner-facing and hiring-facing applications.

• Design, build, and improve speech language models for spoken response understanding, pronunciation analysis, fluency, prosody, and communicative effectiveness.
• Develop and evaluate automated scoring and feedback pipelines for speaking tasks used in:
  • AI-driven speaking practice with instant feedback (learner-facing).
  • Job-relevant oral communication and soft-skills assessments (hiring-facing).
• Train, fine-tune, and evaluate acoustic models, cascading models, speech-to-speech models, speech LMs, and scoring models, including neural and large language model-based approaches.
• Design experiments and conduct quantitative performance, reliability, and validity analyses to ensure assessment quality and decision integrity.
• Work across a range of speaking constructs such as interactional competence, pragmatic competence, spoken critical thinking skills, etc.
• Perform detailed error analysis, intra- and inter-agent rater reliability studies on ASR outputs, spoken features, and scoring behaviors to guide model and product improvements.
• Collaborate with product, UX, and assessment scientists to integrate models into interactive experiences such as practice simulations and hiring workflows.
• Apply responsible AI principles to speech systems, including fairness across accents, dialects, and proficiency levels, as well as transparency of feedback and scores.
• Support model monitoring and governance in production environments, ensuring ongoing quality and compliance for high-stakes use cases.
• Act as the technical lead for an AI conversational assessment product, partnering closely with a Product Manager to translate assessment goals, user needs, and business constraints into model and system design decisions.
• Shape end-to-end conversational assessment design (task structure, prompts, turn-taking, scoring logic, feedback timing) in collaboration with product and assessment stakeholders.
• Balance assessment validity, user experience, system latency, and scalability when making model and system design trade-offs for production conversational assessments.

Qualifications

• Master's or PhD in Computer Science, Electrical Engineering, Speech & Language Processing, Applied Linguistics, Language Assessment, or equivalent applied experience.
• Hands-on experience building speech recognition, spoken language understanding, or automated scoring systems.
• Strong programming skills in Python, with experience using PyTorch or similar ML frameworks for speech and language modeling.
• Solid grounding in machine learning, statistics, and experimental design, especially as applied to model evaluation.
• Experience with modern neural speech models and large language models, including fine-tuning and evaluation for spoken tasks.
• Expertise in model evaluation metrics relevant to speech and assessment (accuracy, reliability, validity, fairness).
• Familiarity with responsible AI practices, including bias analysis, interpretability, and governance for user-impacting systems.
• Strong communication skills, with the ability to explain model behavior and assessment outcomes to technical and non-technical stakeholders.
• Experience working in cross-functional product teams, contributing to roadmap decisions, and shipping ML systems into production.

Nice-to-Have / Domain Alignment

• Experience with spoken feedback systems, pronunciation scoring, fluency analysis, or conversational AI.
• Background in skills assessment, talent evaluation, or hiring platforms using AI-based decision support.
• Familiarity with human-in-the-loop evaluation, rater alignment, or psychometric concepts for AI scoring systems.

Impact of the Role

• Learner-facing speaking practice with immediate, actionable feedback powered by speech LMs.
• Hiring-grade oral communication and skills assessments that support fair, data-driven talent decisions.
• Scalable, responsible speech AI systems that balance technical excellence with assessment validity.