ML Researcher — Speech

Role Overview

This is a full time position, involving the development of Adalat AIs speech recognition transcription product. You will be part of the team in charge of the development and deployment of the Speech-To-Text and Legal Large Language Models

As an early member of the Team, you will: Work closely with founding team to develop the models that power our legal copilot for Judges and Stenographers Identify and drive innovative solutions to address the most critical needs of our Users (Judges at Courts of Different Stages in India). Work in close collaboration with cross-functional partners in design, backend and frontend functions. Solve complex engineering problems for ML Platform. Build cost effective and scalable systems.

Key Responsibilities

Design and implement Hybrid and/or End-to-End speech and language processing system using machine learning and deep learning techniques.

• Preprocess, annotate, and manage large datasets of speech and text for training and testing purposes.
• Develop data augmentation techniques to improve model robustness and generalization.
• Train and fine-tune speech recognition and language understanding models.
• Implement and experiment with state-of-the-art algorithms to optimize model performance.
• Conduct rigorous evaluation of speech models and provide insights for model improvement.
• Collaborate with product managers and engineers to understand user requirements and provide technical solutions.
• Ensure compliance with data privacy and security regulations in projects.
• Stay updated with state-of-the-art techniques in the Speech Recognition field and exchange knowledge with colleagues.
• Coordinate with internal teams to translate business challenges into data pipelines and model frameworks.
• Document research findings, methodologies, and implementation details.
• Communicate progress and results to the team and stakeholders effectively.

Qualifications

Dont worry about ticking all boxes

• 5+ Years working with speech technologies [STT, Diarization, Speech Translation]
• Knowledge with speech technologies like Automatic Speech Recognition (ASR), Multilingual Speech Recognition, Diarization, Language Model, Pronunciation Model, Audio classification.
• Hands-on experience in building Hybrid speech recognition systems like GMM-HMM, DNN-HMM
• Hands-on experience in building end-to-end speech recognition systems using the latest architectures like wav2vec2, Whisper, LAS, Parakeet.
• Familiarity with machine learning and deep learning libraries such as Scikit Learn, TensorFlow, or PyTorch.
• Strong programming experience in languages like Python, C/C++, Shell scripting etc.
• A Bachelors or Masters in Computer Science, Electrical Engineering, or a related field from leading institutes. (a plus)
• Experience contributing to research communities, including publications at conferences and/or journals (a plus)

What You Will Achieve in a Year

Built the ML Stack for Court Systems in India to cater to 5000+ courtrooms running 8-10 hrs per day in the first year

• Take on some of the hardest ML challenges of company, like building feedback loops to improve speech to text performance for regional dialects, legal domain and 10+ Indian Languages.
• Built the best multilingual speech model for legal domain.