Prashaant Mudgala

A Comparative Analysis of LLM Efficacy in Code Generation

Started May 24, 2024

Abstract or project description

The AI landscape is rapidly evolving, with chatbots like ChatGPT and Gemini providing varying strengths and weaknesses in assisting programmers. Still, it remains inconclusive which LLM demonstrates higher performance in code processing tasks in Python and Java, two commonly used programming languages. We hypothesize that ChatGPT would outperform Gemini in generating code snippets for medium and hard difficulty programming prompts in both languages due to its access to a wider range of information (X). To test this hypothesis, data was gathered by using coding prompts from LeetCode as input for both chatbots and evaluating the outputs for correctness, runtime, and memory usage. A total of 50 prompts were used for both ChatGPT and Gemini in Python and Java, and the results were recorded and analyzed. A comparative analysis was performed to determine which LLMs excelled with respect to the two programming languages in terms of runtime and memory usage. The results revealed several applicable takeaways for programmers in industry and students alike. Java prompts on average consumed more memory compared to Python prompts (45.98 MB compared to 14.19 MB), yet yielded shorter runtimes (9.15 ms compared to 331.05 ms). Furthermore, ChatGPT's code solutions on average consumed less memory (29.74 MB compared to 35.57 MB) and had shorter runtimes (141.61 ms compared to 181.65 ms) than Gemini's code. ChatGPT's abilities to generate code snippets exceeds that of Gemini in both aspects of runtime and memory consumption. This indicates that programmers and coders will benefit more from using ChatGPT than Gemini for generating efficient code snippets.

A Historically Accurate Implementation of the Enigma Cipher

Started Aug. 1, 2022

Abstract or project description

Prashaant has shown strong capability for understanding ciphers and cryptographic work in Python. The Enigma cipher, famously used by the Nazis and decrypted mainly by Alan Turing, is a solid project that will teach Prashaant best coding practices for function decomposition and project management. The cipher is complex, and to construct a historically accurate model requires a good understanding of the inner workings of the machine (and some intelligent coding). Along the way, we will learn about the history and techniques used for Enigma encryption and decryption. By the end, we will have a working implementation of the Enigma cipher that anyone can play with!