Apeksha Gaonkar

 |  Email  |  Resume  |  Github  |  LinkedIn  | 


Graduate Student Researcher at Center for Wireless Communications (UCSD).

profile photo

I am a final-year graduate student at UCSD, actively seeking full-time software engineering roles starting in April 2025. With over 2 years of experience in cloud network infrastructure, backend development, and ML infrastructure, I have consistently delivered measurable impact.

As a Graduate Student Researcher, I collaborate with industry leaders like Verizon and Keysight Technologies to advance multi-vehicle perception systems using cutting-edge machine learning and edge computing. My work has improved video stream accuracy by 28% through refined digital twin methodologies. Additionally, I have worked on AI projects such as a Graph Convolution Network-enhanced Retrieval Augmented Generation (RAG) framework for image captioning and a deep learning-based Spotify recommendation system.

My unique blend of software development and machine learning knowledge empowers me to tackle complex challenges and build scalable, high-impact solutions across cloud infrastructure, distributed systems, and backend development. I am eager to leverage these skills to drive meaningful innovations.

Feel free to check out my Resume or reach out via email if you’d like to connect!


 ~  News  |  Experience  |  Education  |  Projects  |  Contact  ~ 


πŸ“… May '24: Presented my research poster and demo at the San Diego Wireless Summit 2024.

πŸ“… Feb '24: Started as a Graduate Student Researcher at Center for Wireless Communications (UCSD).

πŸ“… Sep '23: Started my Master’s in Machine Learning & Distributed Systems at UC San Diego.

πŸ“… Jul '23: Completed 1 year at Arista Networks, focusing on cloud network troubleshooting.

πŸ“… Jul '22: Completed 1 year at Pluribus Networks, developing automated validation frameworks.

CWC @ UCSD logo

Graduate Student Researcher

Center for Wireless Communications (UCSD) | Feb '24 - Present

β€’ Involved in research on enhancing Multi-Vehicle Perception system by optimizing network using Reinforcement learning
β€’ Built an ETL pipeline for KPI sensitivity analysis, processing 1M+ time-series records and identifying key network parameters
β€’ Partnered with Verizon, capturing real-world network data to refine Digital Twin and reduce video stream mismatches by 28%
β€’ Engineered a Multi-Vehicle Digital Twin with sub-1ms latency ZeroMQ messaging, enabling real-time RL training

Arista Networks logo

Software Development Engineer

Arista Networks | Jun '22 - Aug '23

β€’ Built a distributed network observability system in Netvisor OS, enabling real-time pod-to-pod traffic analytics, reducing debug time by 40%.
β€’ Developed an automated certificate rotation system, eliminating 98% of outage risks during upgrades.
β€’ Engineered VXLAN tunneling on RedHat OpenStack, enabling cross-region VM migration.

Pluribus Networks logo

Software Development Engineer

Pluribus Networks | Jul '21 - Jun '22

β€’ Developed PyART, a Python framework, leveraging multithreading and state rollback to optimize Netvisor OS validation
β€’ Collaborated to integrate PyART with the CI/CD pipeline, reducing test execution time by 42% across 35+ automated test suites
β€’ Architected a Netvisor OS monitoring tool using PostgreSQL and FastAPI, identifying 40+ CPU and memory usage bottlenecks
β€’ Built and deployed an ML-powered bug triage system using BERT, XGBoost, Neural Networks, and Weaviate VectorDB, improving classification efficiency by 28%

Pluribus Networks logo

Software Development Intern

Pluribus Networks | Jan '21 - Jun '21

Optimized PostgreSQL queries by 56%, improving bug analysis for regression tests in production builds.

Nischidha Imaging logo

Software Development Intern

Nischidha Imaging | Jun '20 - Dec '20

β€’ Engineered a Flask-based ML backend on AWS EC2 for real-time radiologist assistance, serving 500+ inferences/hour
β€’ Accelerated the inference pipeline by 39% through batching, reducing AWS compute costs by 15%
β€’ Partnered with a fellow intern to reduce false negatives by 13% in deep learning, improving detection accuracy


UCSD logo

University of California - San Diego

Master of Science in Electrical and Computer Engineering

Major: Machine Learning & Data Science

Duration: Sep '23 - Present

Relevant Coursework:
Statistical Learning • Recommender Systems and Data Mining • Machine Learning for Physical Applications • Linear Algebra
Probability and Statistics for Data Science • Introduction to Visual Learning • GPU Programming • Deep Learning

PES University logo

PES University, Bengaluru

Bachelor of Technology in Electronics and Communication Engineering

Specialization: Communication Systems

Duration: Aug '17 - May '21

Relevant Coursework:
Computer Networks • Database systems • Operating Systems • Artificial Neural Networks
Data Structures • Algorithms • C Programming • Python Programming


Graph Convolution Network-enhanced RAG Framework

Achieved a 2.5Γ— improvement in CIDEr scores over baseline models on MS-COCO for zero-shot image captioning.

Spotify Playlist Recommendation

Built a recommendation system using Deep Learning and the Spotify API, processing over 1.5M songs with 30% improved precision.

Distributed Key-Value Store in Go

Developed a distributed key-value store using gRPC, consistent hashing, and RAFT for high availability and fault tolerance.

Optimized CUDA Kernel for GPT-2

Achieved 18% faster execution and 14% lower GPU memory usage by optimizing the CUDA kernel for GPT-2.

California Housing Prices Analysis

Optimized a Random Forest model for price prediction on California housing data (RMSE: 44,437.77).

R-Vine Copula Model for Multimodal Activity Recognition

Designed a novel R-Vine copula model that surpassed state-of-the-art performance by 10.24% (IEEE INDICON 2021).

This template is a modification of Jon Barron's website. Find the source code to my website here.