Shikhar Bharadwaj
Doctoral Student at Carnegie Mellon University
Pittsburgh, Pennsylvania, United States
Summary
👤
Senior🎓
Top SchoolShikhar Bharadwaj is a PhD student at Carnegie Mellon University with a decade of experience building scalable systems and advancing speech and language research, including pre-doctoral roles at Google and DeepMind. His work focuses on integrating phonetic, articulatory, and multimodal cues into large language and multilingual speech models to improve linguistic analysis and representation learning. He brings practical engineering chops—from contributing core data IO improvements to the popular mlpack C++ ML library to enhancing audio encoders in ESPnet—bridging large-scale software design with cutting-edge model development. Comfortable across distributed systems, language modeling, and cross-modal alignment, he pursues foundation-model approaches that capture linguistic structure and variation across languages and speaking conditions. Notably, his background in production engineering informs a pragmatic, scalable approach to experimental design and model implementation.
10 years of coding experience
3 years of employment as a software developer
MTech (Research), Intelligent Systems, MTech (Research), Intelligent Systems at Indian Institute of Science (IISc)BITS Pilani, Birla Institute of Technology and ScienceDoctor of Philosophy - PhD, Doctor of Philosophy - PhD at Carnegie Mellon University
English, Hindi
Github Skills (23)
pytorch10
c-language10
integrate10
armadillo10
machine-learning-models10
machine-learning10
filetype10
file-conversion10
file-operations10
integrations10
file-format10
deep-learning10
fileio10
file-handling10
cprogramming-language10
Programming languages (7)
Github contributions (5)
espnet/espnet
Sep 2024 - Mar 2025
End-to-End Speech Processing Toolkit
Role in this project:
ML Engineer
ML EngineerContributions:38 reviews, 13 PRs, 66 comments in 6 months
Contributions summary:Shikhar's commits primarily involve modifications to the `espnet2/asr/encoder/beats_encoder.py` file, suggesting work related to integrating or modifying the BEATs encoder within the ESPnet framework. These changes include updates to incorporate features like downsampling, conformer adapters, and positional embeddings, indicating a focus on enhancing the audio encoding capabilities of the system. The commits demonstrate an understanding of model architecture, parameter initialization, and the integration of external components (like Wav2Vec2Conformer) into the ESPnet environment.
speech-recognitionspeech-separationchainerspoken-language-understandingspeech-processing
mlpack/mlpack
Dec 2019 - Jan 2020
mlpack: a fast, header-only C++ machine learning library
Role in this project:Back-end Developer
Back-end DeveloperContributions:13 commits, 1 PR, 5 comments in 8 days
Contributions summary:Shikhar primarily contributed to the core data loading and saving functionalities of the mlpack library. They refactored the format guessing code, adding a parameter to specify the data file type for improved flexibility. Their work involved modifying `load_impl.hpp` and `save_impl.hpp` to handle various file formats, including CSV, ASCII, and binary formats, and also included improvements to the file type detection logic. These changes enhance the library's usability and broaden its support for different data sources.
regressionheaderdeep-learningscientific-computingc-plus-plus
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.