Tushar Verma Machine Learning Engineer with 2.5 YOE
No reviews yet

As an experienced machine learning engineer with over 2.5 years of hands-on expertise in the field of Automatic Speech Recognition (ASR) and Language Models, I bring a deep understanding of both the theoretical and practical aspects of these advanced technologies. I have successfully deployed multilingual ASR models, conducted cutting-edge research, and published a paper at InterSpeech 2023, where I proposed innovative methods to improve transcription performance in noisy, code-mixed environments.

What sets me apart as a tutor is my ability to break down complex topics into easily understandable concepts. I believe that every student learns differently, and I take pride in tailoring my teaching methodology to match each student’s unique learning style. Whether it's guiding high school students through foundational concepts or helping undergraduate students grasp more advanced topics, I always emphasize clarity, step-by-step problem solving, and real-world application.

My teaching process involves:

Hands-on Learning: I encourage students to actively engage with the material by applying what they've learned to practical problems. This fosters a deeper understanding and builds confidence in applying these concepts independently.
Clear Explanations: With experience working on state-of-the-art ASR models like wav2vec 2.0 and Whisper, I ensure my lessons are not only technically sound but also accessible to learners of all levels.
Real-World Applications: Having developed solutions that are used in business analytics and multilingual transcription, I relate the theory to real-world use cases, helping students see the value of what they are learning beyond the classroom.
Results-Oriented Approach: I’ve trained models that achieved significant improvements in word error rate, and I bring the same results-driven mindset to teaching. My students can expect to see measurable progress in their understanding and performance.
Through personalized attention and a passion for making complex topics accessible, I aim to help students excel in their studies and build a solid foundation for future success.

Subjects

  • Machine Learning Grade 12-Bachelors/Undergraduate

  • Deep learning with Python programming Grade 12-Bachelors/Undergraduate

  • Python 3 Grade 12-Bachelors/Undergraduate


Experience

  • Machine Learning Engineer (Feb, 2022Present) at Convin, Bengaluru
    Developed and deployed multilingual ASR models supporting Hindi, English, Kannada, and Tamil, achieving an average Word Error Rate (WER) of 21, with the best model reaching 15 WER.
    Published research on ASR for low-resource and multilingual code-mixed speech at InterSpeech 2023, proposing a novel method for training wav2vec 2.0 models.
    Experimented with SOTA ASR models (wav2vec 2.0, Whisper, WavLM) and language models (KenLM, awd-lstm, transformers) using PyTorch, HuggingFace, Fairseq, and S3PRL toolkits.
    Created a pipeline for rapid modification of Language Models to include domain-specific vocabulary, enhancing business analytics.
    Developed POCs for MLOps using MLflow and data infrastructure using Delta Lake, improving model versioning and data reliability.
    Skilled in data curation, web scraping, and exploratory data analysis for ASR and language modeling tasks.
    Familiar with GCP and Linux environments.

Education

  • Bachelors of Tecnology Computer Science (Apr, 2016Apr, 2020) from Dr. A. P. J. Abdul Kalam Technical University, Lucknowscored 7.8 GCPA

Fee details

    300500/hour (US$3.545.90/hour)

    Subject complexity and level play a significant role in determining the fee, with college-level or specialized topics generally commanding higher rates than high school subjects. The student's specific learning needs, such as exam preparation or remedial support, are also factored into the pricing. Time slots during peak hours may incur slightly higher fees due to increased demand. However, we offer package discounts for regular, frequent sessions to encourage consistent learning. To provide comprehensive support, all fee segments include homework assistance and test creation to a reasonable extent, ensuring students receive well-rounded academic support without incurring additional charges for these essential services.


Reviews

No reviews yet. Be the first one to review this tutor.