AI Engineer – ASR

This is a fully remote position, open to applicants in India.

📋 Description

• Lead the design and implementation of speech recognition systems for automated reading assessments in various languages.

• Engage with the entire ASR pipeline, including data collection, annotation, preparation, model training, decoding, alignment, error analysis, and deployment.

• Provide technical guidance for speech data collection across diverse languages, dialects, age groups, reading levels, environments, devices, and target items.

• Establish annotation guidelines and quality assurance processes for transcripts, reading errors, disfluencies, background speech, and related phenomena.

• Identify data deficiencies and inconsistencies in annotations through model error analysis.

• Suggest focused data collection or re-annotation strategies.

• Design annotation frameworks and inter-annotator quality evaluation processes.

• Train, fine-tune, and assess ASR models for children's reading speech across different languages and accents.

• Create pipelines for audio preprocessing, segmentation, normalization, transcription alignment, and model training.

• Modify ASR architectures such as Whisper, wav2vec 2.0, Conformer, CTC/RNN-T, or similar models.

• Optimize models for low-bandwidth, store-and-forward, server-based, and potentially offline/mobile inference scenarios.

• Identify substitutions, omissions, insertions, repetitions, partial-word attempts, hesitations, and self-corrections.

• Implement forced alignment, phoneme-level alignment, or pronunciation-scoring methods as needed.

• Conduct systematic error analysis based on language, reading level, age, accent, gender, recording conditions, and device usage.

• Collaborate with assessment specialists and field teams to translate model outputs into metrics such as reading accuracy, WCPM, fluency, and reading-level assessments.

• Define and monitor ASR metrics including WER/CER, word-level accuracy, error-detection precision/recall, alignment accuracy, and agreement of reading scores with human evaluators.

• Establish reproducible experimentation and model evaluation workflows.

• Maintain comprehensive technical documentation.

⛳️ Requirements

• A Degree in Mathematics, Computer Science, Data Science, or another related STEM field.

• 3-5 years of full-time work experience; at least 2 years in an AI/ML position.

• Extensive experience in automatic speech recognition, speech processing, or spoken-language technologies.

• Proficient in Python and experienced with PyTorch or similar deep-learning frameworks.

• Familiarity with modern ASR architectures and toolkits such as Hugging Face Transformers, NeMo, Kaldi, or equivalent.

• Knowledge of acoustic modeling.

• Understanding of CTC, transducer, and encoder-decoder architectures.

• Familiarity with decoding and beam search techniques.

• Understanding of language models.

• Knowledge of forced alignment methodologies.

• Familiarity with phoneme representations.

• Understanding of WER/CER and ASR error analysis techniques.

• Experience in preparing and handling large-scale audio and transcription datasets.

• Ability to systematically diagnose model errors rather than solely relying on aggregate transcription accuracy.

🏝️ Benefits

• Flexible, remote-first work environment.

• Work from any location within India.

• Periodic travel to field locations.

• Freedom to work from wherever you feel most productive.

• Opportunity to observe models in action and collect real-world feedback.

• Competitive compensation aligned with industry standards and commensurate with experience.

• Contribute to a global impact by supporting the learning journeys of millions of children.

• Tackle unique technical challenges in AI/ML, ASR, and low-bandwidth/offline environments.

• Freedom to experiment with the latest state-of-the-art models.

• Collaborate with a global, mission-driven team.

• Enjoy an inclusive and collaborative work environment.

• Equal opportunity employer.