*

here are some of the projects i've worked on.

Ad-Astra

xAI Hackathon Project (Dec 2025)

Ad-Astra personalizes ads on X by learning a user’s interests and writing ads in their style, then estimating which ads are most likely to get clicks using Grok models.

achievements

  • Demoed this project at the xAI Hackathon 2025; selected for an all-expenses-paid presentation at xAI headquarters in Palo Alto.
  • Built an end-to-end system that analyzes X activity (posts, likes, bookmarks, timelines), creates persona-based ad variations, and scores them for click likelihood before delivery.
  • Built a residual attention model as a part of the pipeline that predicts when and where an ad should appear in-feed enabling higher CTRs.

technologies

pythongrok modelsxai apisupabaseflaskreactnext.js

OpenProbe

creator (may 2025 - june 2025)

an open-source deep research agent using python, langchain, and langgraph. designed to outperform existing search and research systems through advanced multi-hop reasoning capabilities.

achievements

  • outperformed openai's gpt-4o-search on the FRAMES benchmark (+1.5% accuracy)
  • state-based orchestration of agents/tools like websearch, coding and logical reasoning via LangGraph.

technologies

pythonlanggraphagentic ailambda apitypescriptgemini api

PersonaAI

creator (Oct 2025)

Build and test A/B digital campaigns using millions of AI-generated personas. PersonaAI automates persona creation and variant generation to help marketers maximize engagement and conversion.

achievements

  • Won 3rd place at the lovable.dev hackathon, created this in 6 hours.
  • Still in development, but will be released soon.

technologies

pythonmicrosoft phi-4gemini apitypescriptlovable.dev

KeyCognition

DataBricks Hackathon Project (Nov 2025)

Predicting cognitive load of users based on their behavior and interactions with their keyboard stroke data using machine learning models.

achievements

  • Won 3rd place at the UW Databricks Hackathon, created this in 12 hours.
  • Utilized 136 million+ keyboard stroke data to train the machine learning models.

technologies

pythonpandasnumpyscikit-learndatabricksazureazure mlazure sql

Detecting GAN Generated Deepfake Images

creator (jan 2021 - june 2021)

research project focused on detecting gan-generated deepfake images using custom convolutional neural network architectures. achieved state-of-the-art performance on the stylegan dataset.

achievements

  • achieved 97.77% accuracy on the StyleGAN dataset
  • designed custom cnn architecture optimized for deepfake detection

technologies

pythontensorflowkerasdeep learningcnncomputer vision

Plant-FATE

contributor (Jan 2022 - Aug 2022)

Plant-FATE is an eco-evolutionary vegetation model that accounts for multi-timescale adaptations of invdividual plants and plant species to the environment.

achievements

  • Contributed to software development for the Plant-FATE simulation model, developed key modules in C++.
  • integrated forest patch models to simulate plant responses in climate variations, analyzed generated plots using R programming.

technologies

c++r