Abel Yagubyan | Software & ML Engineer

Hi, I'm Abel Yagubyan.

A

Self-driven, quick starter, passionate programmer with a curious mind who enjoys solving challenging real-world problems.

About

Welcome to my page! I am a Bachelor's graduate in Computer Science and Applied Mathematics from UC Berkeley. I have a particular passion for Machine Learning, Distributed Systems, Databases, and Security. Building on this passion, I pursued and successfully completed my Master's of Science degree in Computer Science at Northwestern University. I have previously worked on technologies like Python, C/C++, Golang, Java, MySQL, MongoDB, Flask, and x86/RISC-V during my Bachelor's/Master's. I am passionate about developing complex applications that solve real-world problems impacting millions of users within my intended areas of interest.

Languages: Python, Java, C/C++, Golang, Rust, Swift, PHP, JavaScript, Bash
Databases: MySQL, PostgreSQL, MongoDB, Firebase
Libraries/Frameworks: Keras, TensorFlow, PyTorch, CUDA, UPC++, GPU Programming, MPI, NumPy, Pandas, OpenCV, Flask, Django, Node.js
Tools & Technologies: Git, Docker, AWS, Heroku

Looking for an opportunity to work in a challenging position combining my skills in Software & ML Engineering, which provides professional development, interesting experiences, and personal growth.

Experience

Northwestern University’s Prof. Guo

Research Assistant

Contributor to Professor Dongning Guo's and PhD student Yiming Zhang’s project titled as "Deep Reinforcement Learning for Scheduling, Power Control, and Rate Adaptation in Cellular Networks".
Currently speeding up project simulations using GPU programming.
Tools: Python, GPU programming, and CUDA

June 2022 - April 2023 | Evanston, IL

Lawrence Berkeley National Laboratory

Research Assistant

Working with the Pagoda Project team at Lawrence Berkeley National Laboratory on UPC++, a C++ library providing classes and functions that support Partitioned Global Address Space (PGAS) programming, to:
Develop and deploy automated performance regression testing for UPC++ using crons and Python.
Implement benchmarks in UPC++ (OSU micro-benchmarks, Parallel Research Kernels, Bale benchmarks) and compare to other libraries such as MPI and SHMEM.

Tools:

June 2022 - April 2023 | Berkeley, CA

UC Berkeley's Prof. Filippenko

Research Assistant

Worked as a Research Assistant for the world-renowned Professor Alex Filippenko to implement an ML Classification model using CNNs that use image data from Caltech's ZTF program to accurately classify different types of supernovae and transient objects.
More mentioned in the Projects section.
Tools: Python, JS

Aug 2021 - May 2022 | Berkeley, CA

UC Berkeley's CS61C and CS161 Courses

Teaching Assistant

Teaching Assistant for CS61C: Computer Architecture and CS161: Computer Security courses.
Project lead for CS61C's numc, a project that implements a simple numpy library and achieves up to 20x speedup (in comparison to dummy implementation) with the help of OpenMP, SIMD, and Loop unrolling.
Project lead for CS61C's CS61Classify, a project that implements a simple Neural Network model using RISC-V assembly code (similar to x86).
Project lead for CS161's Dropbox project, a project that implements an encrypted, safe, and scalable file sharing system using Go along with Cryptography.
Tools: Python, C, Go, Assembly (RISC-V and x86), JS

May 2021 - May 2022 | Berkeley, CA

Apple x UC Berkeley

Software Engineer Intern

In collaboration with Apple and UC Berkeley's CS61C: Computer Architecture course (myself as a Teaching Assistant for that course), introduced Computer Architecture-based projects to 4000+ HBCU students by implementing the source code with the help of C, Python, and Assembly.
Project lead for numc, a project that implements a simple numpy library and achieves up to 20x speedup (in comparison to dummy implementation) with the help of OpenMP, SIMD, and Loop unrolling.
Project lead for CS61Classify, a project that implements a simple Neural Network model using RISC-V assembly code (similar to x86).
Tools: Python, C, Assembly (RISC-V)

May 2021 - Aug 2021 | Remote

UC Berkeley's Prof. Garcia

Research Assistant

Introduced new online examination features to 3,000+ students with 100% satisfaction by implementing an interactive RISC-V compiler to PrairieLearn, an online examination platform, for UC Berkeley Computer Architecture courses using a combination of Python, Assembly (RISC-V), C, JS, and Docker.
Published paper with Prof. Garcia for the SIGCSE '22 Technical Symposium.
Tools: Python, Docker, Assembly (RISC-V), C, JS

June 2019 - May 2022 | Berkeley, CA

Projects

                
FibonAI
                  The premier legal AI-powered workspace that transforms in-house legal operations
                
AccomplishmentsTools: Python, Flask, AWS, SQL, Zilliz, and JS
End-to-end in-house legal workflow with GPT-powered suite of tools.
Onboarded over 10 beta customers from enterprises like Confluent and Cresta.
Helped save up to 40% of repetitive and manual daily tasks in in-house teams.

SCALPEL
                  Introducing efficiency in Deep Neural Networks
                
AccomplishmentsTools: Python, C++, Bash
Reduced popular modern Deep Neural Network (DNN) model sizes (i.e., AlexNet, LeNet-5, etc.) by up to~80%.
Achieved up to ~3.5x speedup in modern DNNs.
Customizes DNN pruning techniques to the underlying hardware parallelism (Microcontroller, CPU, and GPU) 

Supernovae Classification
                  Supernovae type classifier using Convolutional Neural Networks (CNNs) and Random Forests
                
AccomplishmentsTools: Python, Flash, Bash
Retrieves image data from Caltech's ZTF Program API.
Processes and classifies Supernovae according to their type by up to ~95% accuracy by using a modified version of ResNet-18.