_ANDREJ_KARPATHY_LECTURES_
[CURRENT]
Currently working through Karpathy's YouTube lectures on transformers to build my own nano-GPT. Code to come soon.
[2023 - CURRENT]
Created a semi-automated trading system that uses Bollinger Bands to identify trading opportunities for different securities. Also built a companion library to analyze portfolio results & conduct backtests. The trading algorithm runs daily & sends trade notifications promptly after market-close. You can explore the code in a live environment - check out research/Example.ipynb here.
[2024]
Built the game "snake" for a micro:bit V1 controller to learn about using Rust for embedded development. Walked away with a better understanding of crates, scope/borrowing, and a compiler that actually throws useful errors.
[2022]
Used Monte-Carlo simulation techniques to identify the most lucrative property in Monopoly.
[2021]
Developed an autocomplete tool to help medical scribes quickly note patient/MD conversations. Scraped medical terminology from a textbook and built a custom trie structure to load the glossary & facilitate quick autocomplete.
[2020]
Created backend system for an application designed to replicate Vannevar Bush’s “memex.” Built fully-featured backend to handle user registration, account deletion, login, search, filtering via tags, along with basic CRUD operations on memexes.
[2020]
Created a conversational virtual assistant with custom integrations - paying/requesting close friends through Venmo, handling Spotify sessions (pause, play, skip, creating a new playlist), sending texts, and summarizing search queries. Pre-ChatGPT.
[2020]
Capstone Project involved designing a fully functioning search engine - parser, crawler, index, constraint solver, query compiler, and a ranker. This was built entirely in C++ and because using the STL was prohibited, elements such as vectors, arrays, hashtables, and even strings were built from scratch. Repository is currently private.
[2019]
Backend-only hackweek project that allowed users to order Lyfts via SMS. Designed for communites with low smartphone adoption.