Atharva Jadhav profile picture

I am Atharva Jadhav, a developer from Mumbai, India.

Scroll down

Qualifications

Master of Science in Computer Science Engineering

University at Buffalo

GPA: 3.81

Year: 2024-2026

Transcript:

Master of Science in Computer Application

Symbiosis Institute of Computer Studies & Research

CGPA: 8.83

Year: 2020-2022

Transcript:

Bachelor of Computer Application

Symbiosis Institute of Computer Studies & Research

CGPA: 7.93

Year: 2017-2020

Transcript:

Work Experience

Research Assistant

University at Buffalo

March 2025 - Present
  • Visualized acoustic word clusters in 3D by projecting wav2vec2 audio embeddings with t-SNE and aligning text using the Montreal Forced Aligner (MFA).
  • Investigating duration reduction mechanisms in state-of-the-art automatic speech recognition (ASR) and text-to-speech (TTS) systems, focusing on homophones and lexical frequency.
  • Devising methods to generate FOILS for a Maze app, enabling users to learn low-resource languages with ease.
  • Replicating the experiments of 'BERT Rediscovers the Classical NLP Pipeline' with Pythia Model.

Research

Multi-Task Learning for Low-Resource Speech Emotion Recognition

Dr. Nasrin Akhter

Developing a Multi-Task Learning (MTL) model for low-resource Speech Emotion Recognition (SER) by leveraging ASR and multilingual datasets to distill powerful linguistic and prosodic features.

Duration Reduction in ASR/TTS Systems with Acoustic Analysis

Prof. Cassandra Jacobs

Investigated duration reduction in ASR/TTS systems by analyzing homophone and lexical frequency effects, and developed 3D visualizations of acoustic word clusters using wav2vec2, t-SNE, and the Montreal Forced Aligner (MFA).

Personal and Academic Projects

Independent Study (Thesis advisor: Dr. Nasrin Akhter) (Dec 2024 – Present)

• Architecting a generalized multilingual SER model that rivals specialized systems, targeting an average performance within 20% of single-language SOTA models, by fine-tuning Wav2Vec2 with a multi-task ASR approach.

Augmentative and alternative communication for societal good (Mar 2025 – May 2025)

• Created an efficient, on-device conversational assistant for AAC users, packaged into a 5GB deployable model that runs without an internet connection, by fine-tuning and quantizing LLMs like LLAMA3-8B for local inference.

Traffic light automation using RL to facilitate emergency vehicles (Jan 2025 – Apr 2025)

• Minimized emergency vehicle delay at intersections, reducing average wait times by 45% compared to a standard timed system, by implementing and comparing a suite of RL algorithms (Q-Learning to DDQN) in the SumoRL environment.

Wikipedia Chat Bot (Oct 2024 – Dec 2024)

• Built a scalable search and summarization system by scraping 50,000 Wikipedia summaries, indexing them with SOLR, and deploying a Flask server with a React frontend on GCP.
• Designed an intelligent response pipeline using zero-shot classification for message categorization, integrating a Blenderbot-based chatbot for casual conversations and T5 for summarizing SOLR query results.

In-place convolution with OpenMP (Oct 2024)

• Developed a C++ program applying a 3 × 3 matrix kernel to a 1D vector (representing a 2D float array) using multithreading with OpenMP, achieving 70% efficiency on 64 processors in an HPC environment.

Visualization of WHO. data

Visualizes 618 datasets from WHO API in line-charts dynamically, built in latest NextJS 13 framework.

Certificate & Courses

  • Machine Learning by Andrew Ng (Coursera & Stanford online).
  • The complete 2021 Web development Bootcamp by Angela Yu (UDEMY).
  • Architecting with Google Compute Engine. (Coursera & Qwiklabs).
  • Participation in Symtaxify Hackathon. (ACM).

Lets get in touch