Mihir Pavuskar
Software Developer
Click me!
Seattle, WA
01
About
Currently pursuing MS in Computer Science (Artificial Intelligence) at the University of Southern California with a GPA of 4.0. Have extensive expertise in Machine Learning, Artificial Intelligence, and Software Development. Have work experience as SDE (React), Researcher (NLP). Have built several projects using technologies like Computer Vision, NLP, Tensorflow, PyTorch, Reach, NodeJS, etc.
02
Experience
Aug 2024 - present
- Lead scalable microservices architecture in Go using gRPC framework integrating PostgreSQL, S3, and AWS Lambda, reducing system latency by 40% while ensuring GDPR compliance across 100,000+ user records.
- Implement comprehensive testing framework with unit, integration, and chaos testing, achieving 95% test coverage.
- Develop and fine-tune LLM and RAG systems using PyTorch for automated campaign content generation, resulting in 25% increase in user acquisition and streamlined marketing operations.
- Implemented sophisticated PII redaction mechanisms and synthetic data generation for LLM training and inference, ensuring GDPR compliance while maintaining model performance across 10,000+ user records
Sep 2022 - Dec 2024
- Developed critical features in video/audio editing studio using React and TypeScript, implementing custom hooks and context providers that improved code reusability by 35% and reduced bundle size.
- Analyzed Google Analytics logs to identify and resolve bugs and devise fail-safes reducing crash rates by 40%.
- Engineered performance optimizations through UI virtualization and Redux state management patterns, resulting in 5x faster rendering for data-intensive media projects and 60% reduction in memory usage.
- Collaborated with Customer Success team to upgrade Google Analytics to GA4 while maintaining data consistency.
Feb 2022 - Jul 2022
- Architected end-to-end financial question-answering system using ensemble of BERT-based models (TaPaS, ALBERT, HybridR), achieving state-of-the-art accuracy of 72% on FinQA benchmark dataset.
- Engineered custom data preprocessing methodology integrating structured and unstructured financial data processing over 1,000 financial documents with 99.7% accuracy
Jan 2022 - Apr 2022
- Overhauled dashboard by migrating from legacy code to NextJS with modular frontend components achieving 30% code reusability and implementing lazy loading that reduced initial load time by 40%.
- Engineered real-time ECG monitoring system using NextJS with custom ReactCanvas visualization library, implementing WebSocket for live data streaming and reducing latency by 60% while handling 50+ concurrent patient sessions.
Jul 2020 - Mar 2021
- Designed intelligent text normalization system for ASR using ensemble of BERT and XGBoost models, improving model accuracy by 4.4% on production data.
- Engineered end-to-end data pipeline for slang normalization, with custom preprocessing algorithms for handling non-standard words and internet slang, achieving 98.75% classification accuracy across 17 test batches of 500,000 samples each.
03
Education
University of Southern California
Master of Science in Computer Science
Jan 2023 - Dec 2024
4.0
GPA
Vellore Institute of Technology
B.Tech. in Computer Science and Engineering
May 2018 - Jul 2022
3.8
GPA
04
Project
Resume AI
May 2023 - Aug 2023
- Implemented full-stack resume generation platform using NextJS and Vercel-AI SDK, implementing server-side streaming that reduced response latency by 60% while handling 1000+ concurrent user
Intelligent Pente Playing Agent
Jan 2023 - May 2023
- An Intelligent agent which plays the game Pente, a stratergy board game played on a 19x19 board. Built using C++ and utilizes Alpha-Beta Pruning
Smart Checkout System
Aug 2020 - Jun 2021
- A smart checkout system inspired by the Amazon Go store, coupled with inventory management and a sales prediction system.
See Food
Oct 2020 - Dec 2020
- Shazam for Food, can identify and classify up to 500 dishes using uploaded or live pictures
05
Certification / Achievements
Deep Learning Specialization
Coursera - Andrew Ng
Date Issued: Jun 2020
Credentials: QMV95NA36A3W
AWS - Cloud Practitioner
Amazon Web Services
Date Issued: August 2021
Date Expired: August 2024
Credentials: EBT4G4WJEMV117GB
Python Data Visualization
Coursera - Rice University
Date Issued: Jun 2020
Credentials: FGJQREHKDTB6
NOS - Security Analyst
NASSCOM
Date Issued: Jan
07
Skills
08
Hobbies
09
Contact