My Projects

From architecting high-throughput cloud platforms to implementing complex deep learning models, I specialize in building robust, data-intensive applications. Below is a selection of my work.

software

Cloud-Native ETL & Analytics Platform

Carnegie Mellon University

Feb 2025 - Apr 2025

AWS (EKS)Go (GoFiber)SparkScalaKubernetesRedisElastiCacheTerraform
  • Engineered Go (GoFiber) microservices on Kubernetes that handled ~125K RPS with a P99 latency of 5ms.
  • Built a cost-optimized ETL pipeline using Spark (Scala) to process 1TB of raw data from Azure Blob.
  • Secured 3rd place out of 40 teams by designing a high-throughput, low-latency data platform on AWS.
software

Real-Time Stream Processing System

Carnegie Mellon University

April 2025

KafkaSamzaYARNJava
  • Designed a stream-processing system using Kafka and Samza to join multiple real-time GPS data streams.
  • Deployed the driver-matching application on a YARN cluster, ensuring high availability and low-latency processing.
ai ml

Generative Diffusion Models for Image Synthesis

Carnegie Mellon University

Sept 2024 - Dec 2024

PyTorchPythonU-NetVAEDDPM/DDIM
  • Implemented Denoising Diffusion Probabilistic Models (DDPM) and a U-Net architecture from scratch in PyTorch for iterative image generation.
  • Integrated Variational Autoencoders (VAE) and Classifier-Free Guidance (CFG) to enhance conditional generation, evaluating performance with FID scores.
software

Jira Sprint Analytics Dashboard

Personal Project

May - Jun 2025

JavaScriptHTML/CSSChart.jsJira REST API
  • Developed a client-side web dashboard using vanilla JavaScript and Chart.js to visualize real-time sprint data from the Jira REST API.
  • Provides teams with actionable insights into velocity, scope changes, and task completion trends to improve agile workflows.
Live Demo View Code
software

Heterogeneous Cloud Storage System

Carnegie Mellon University

March 2025

MySQLMongoDBRedisTerraformAzureJava
  • Designed a polyglot persistence solution for a social network application, using the best database for each specific data type.
  • Utilized MySQL for structured data, MongoDB for flexible user profiles, and an in-memory Redis caching layer to accelerate queries.
ai ml

Attention-Based Speech Recognition Transformer

Carnegie Mellon University

Nov 2024

PyTorchPythonTransformersAttention Mechanisms
  • Built an end-to-end, attention-based speech-to-text transformer from scratch.
  • Implemented the full encoder-decoder architecture with multi-head self-attention and cross-attention mechanisms in PyTorch.
ai ml

Sequence-to-Sequence Phoneme Transcription

Carnegie Mellon University

Oct 2024

PyTorchPythonLSTMs/GRUsCTC Loss
  • Developed a deep sequence-to-sequence model using LSTMs to map audio utterances to phoneme sequences.
  • Implemented Connectionist Temporal Classification (CTC) loss to handle alignment challenges between variable-length audio and text.
ai ml

Human Facial Expression Classification

BITS Pilani

Feb 2021 - Apr 2021

PythonCNNsscikit-learnOpenCV
  • Implemented and evaluated various CNN architectures for facial expression recognition.
  • Developed strategies to handle class imbalances and overfitting to achieve competitive evaluation metrics on a public dataset.