Aarush Verulkar
Writing clean, efficient code and building AI solutions.
I'm a goal-oriented student with experience in Machine Learning, Blockchain, and Full Stack Development. I've worked for startups and enterprises, and enjoy writing new code as well as contributing to open-source projects. I'm optimistic and business-minded, and consider collaboration and communication strong suits.
I graduated from Mumbai University with a bachelor's in Computer Engineering and I am currently pursuing my master's degree in Computer Science at USC. I try to keep up with data structures and algorithms in my free time, and routinely explore ML. When I'm not coding, I'm either outside, traveling, cooking, playing Table Tennis, or reading a history book. I'm also passionate about startups, F1, anime and everything tech. I'm usually wearing a smile.
Projects
Lumiere: XAI Movie Recommender
Explainable AI Movie Recommendations powered by Knowledge Graphs. Uses Node2Vec embeddings and Neo4j to provide not just recommendations, but clear explanations of why you will love each movie based on actors, directors, and genres.
Intellihealth: Intelligent Health Record System
A blockchain-based electronic health record system ensuring secure digital storage and enhancing patient data accessibility.
PocketLLM
A quantized large language model optimized for deployment on resource-constrained devices, implementing efficient model compression techniques for edge computing applications.
Experience
May 2025 - Sep 2025
Research Software Engineer
University of Southern California
- Overcame data ingestion bottlenecks for 26K+ daily records by architecting distributed Apache Airflow DAGs with parallel task execution, reducing pipeline latency by 60% while ensuring strict ACID compliance for research data.
- Engineered high-throughput microservices using FastAPI and AsyncIO, implementing connection pooling to handle 100+ concurrent requests while maintaining sub-50ms P99 latency during peak load.
- Accelerated deployment cycles by constructing a CI/CD pipeline using GitHub Actions, automating unit/integration testing to cut manual release time by 20% and eliminate environment-specific bugs.
- Diagnosed critical CPU bottlenecks using cProfile, optimizing expensive object allocations and loop overheads to reduce server resource consumption by 47%.
Jun 2023 - Jul 2023
Software Engineering Intern
BMH Technologies
- Architected event-driven Python microservices to process high-velocity telemetry data, utilizing Redis for message buffering to achieve <200ms end-to-end latency and improving anomaly detection accuracy by 25%.
- Eliminated 40+ hours/week of manual operational overhead by engineering robust SQL ETL pipelines, automating complex joins and validation across 12 tables with zero data integrity loss.
- Developed interactive monitoring dashboards using React and Plotly, rendering real-time operational metrics from PostgreSQL to reduce incident response time by 30% for the operations team.