2020 - 2021

Master's degree (MS) in Machine Learning

at San Jose State University, California, USA

Hands-on experience with almost all traditional ML algorithms/techniques(like one hot ecnoding, scaling, Normalization, PCA), libraries (infra ecosystem) and architectures, having applied them to multiple datasets.

Explored the evolution/timeline of AI, from the early days of the perceptron to Hopfield and Boltzmann networks, followed by Recurrent Neural Networks (RNNs), Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and the more recent developments in attention mechanisms. Realized history of Computing is just history of creating Artificial Intelligence and Von Neumann architecture might just be a detour.

I tried to unlock GPT-2, but apparently, it was too powerful for mere mortals like me at that time. So, I had to make do with the ancient arts: unigrams, bigrams (Tf-idf), beginner-level tokenizing, basic NLP spells, Word2Vec, and the legendary ImageNet/WordNet combo. Basically, I was stuck in the Stone Age while GPT-2 laughed from its ivory tower

I dove into the wild world of Diffusion models, Transformers, and GANs. Basically, I spent my time making machines dream, transform, and battle in pixelated arenas. Fun times in the AI dungeon

Few Projects

• Recommendation System With Compressed Bit vectors (Bit slice index arithmetic):
  • Implemented Enhanced Word Aligned Hybrid (EWAH) compression in the recommendation system to enhance space efficiency and accelerate performance, effectively handling sparse and dense bitmaps, thereby optimizing storage overhead and query processing speeds.
  • Compared the performance with traditional RecSys techniques like Collaborative Filtering, Matrix Factorization (SVD, SVD++, Non-negative Matrix Factorization)
• Yelp Prototype (React, Node, MySQL, MongoDB, Redux, JMeter, Kafka, AWS):
  • Developed a clone of yelp restaurant website and improved database performance by 25% using MySQL connection pooling to enable reuse of multithreaded connections.
  • Increased the scalability of the system by kafka streaming and deploying the app on EC2 and database on RDS.
  • Implemented State management, tested backend in JMeter & Mocha for 10000 concurrent users & frontend in Enzyme.
• Point clouds visualization and comparison with python pptk and open3d libraries RealTime_PointClouds_Visualization
• HumanActivityDetectionWithSmartPhones Human Activity Detection sensor data analysis and visualization to find correlations between activities and plot the high dimensional latent space data using t-sne.
• Used Appium framework and jmeter to test a AI therapist Mental Health chatbot called Wysa: jmeter appium
• Twitter covid19 tweets sentiment analysis using twitter api, nltk and textblob libraries

Coursework

• Enterprise Distributed Systems
• System Software
• Advanced Computer Architecture (MIPS, ARM (CISV vs RISC basics))
• Advanced Microprocessor Design (Developed a graphics engine using Liquid Color Display on LPC1769, a ARM Cortex-M3 microcontroller. Utilized an 18-bit graphical LCD for displaying various shapes and patterns via SPI communication with the LPC1769 module.)
• ML:
• Advanced Data Mining
• Testing Intelligent Systems (AI model benchmarking and evaluation techniques)
• Business Intelligence
• Machine Learning
• Deep Learning (Neural Networks)
• Recommendation Systems (Collaborative Filtering and Matrix Factorization)

hello world

def hello_world():
    print("Hello, AI World!")

hello_world()