×
Home About Blog Favorite Song Experience Resume CV Contact
☰

Projects

Portfolio Predictor

Quantitative Stock Forecasting Using ML Models

  • Details:
    • Financial time-series forecasting using 10 years of market data
    • Random Forest, XGBoost, LSTM, etc predictive models
    • MPI-parallelized training and evaluation pipeline
    • Backtesting engine with Sharpe ratio and drawdown analysis
    • Languages: Python
    • Tools: PyTorch, SciKit-Learn, etc
  • GitHub

KlaudOS

Klaud themed Operating System

  • Details:
    • Handmade 2 stage Bootloader and Kernel
    • Klaud file system
    • 32 bit protected mode
    • Interactive shell
    • Languages: C, x86 assembly
  • GitHub
  • Demo Video

40kAI

Using Reinforcement Learning to play Warhammer 40k

  • Details:
    • DQN using PyTorch
    • Custom Warhammer Gymnasium Environment
    • Interactive GTK Application used to train, evaluate, and play against the model
    • Web Scraper to retrieve data from Wahapedia
    • Languages: Python, C++, C
    • Tools: PyTorch, Gymnasium, gtkmm
  • GitHub

KlaudSynth

Using DSP to make my own command line synthesizer

  • Details:
    • DQN using PyTorch
    • Real-time audio synthesis supporting sine, sawtooth, triangle, and square waveforms
    • Implemented 8 different filters
    • Designed ADSR envelope system
    • Implemented 6 effects
    • Developed accompanying synthesizer PCB schematic for hardware implementation
    • Languages: C, Python
  • GitHub

Obsidian-kak

Kakoune Plugin for Obsidian Integration

  • Details:
    • Made for those who love Obsidian, but also love their Kakoune config
    • Languages: C, KakouneScript
  • GitHub

SolvedCrackMes

Repository of CrackMes and their solutions (For Linux)

  • Details:
    • CrackMes from crackmes.one, challenges.re, NoraCodes
    • Each folder has an ELF file and C file with deduced source code
    • explain.txt for every problem that explains every solution
    • Languages: C & Python
    • Tools: Radare2, Ghidra, Objdump
  • GitHub

Number Station Id

Using Machine Learning to Identify Number Stations

  • Details:
    • CNN (built from scratch) used for classification
    • Interactive GUI made with PyQt5
    • Can identify up to 15 different number stations
    • Languages: Python
  • GitHub

MMBCReco

Using Machine Learning to Identify/Classify tracks in a bubble chamber

  • Details:
    • DCGAN that generates data from a video of a Bubble Chamber
    • Faster R-CNN that can detect tracks
    • Script that can produce track data with annotations in COCO format
    • Languages: Python
    • Tools: PyTorch & Detectron2
  • GitHub