|
Sid Ijju I am a final-semester Master's student in Artificial Intelligence at Columbia University. Broadly, I am interested in using deep learning to build intelligent agents that can reason about the world and collaborate with humans. I'm also deeply invested in working on multimodal applications of generative A.I. Industry Experience. I previously interned at Stripe, Cohere, Amazon, Quantel AI, and Correlia Biosystems. Other Experience. I completed my undergrad at U.C. Berkeley, where I was part of Berkeley's Management, Entrepreneurship, and Technology (M.E.T.) Program. I was the Head TA for Berkeley's CS 188 (Artificial Intelligence), and a member of Machine Learning at Berkeley (ML@B) and Voyager Consulting. |
|
Industry Experience |
|
Stripe
Software Engineer Worked on enumerating and providing visibility into unmodeled network fee costs. |
|
Cohere
Software Engineer Worked on multimodal data acquisition and formatting for instruction fine tuning. |
|
Amazon
Software Engineer Worked on device-agnostic command line interfaces across the AWS network. |
|
Quantel AI
Machine Learning Engineer Worked on feature engineering for financial analysis and yield curve prediction algorithms. |
|
Correlia Biosystems
Software Engineer Worked on computer vision and pattern analysis algorithms for biological assay analysis. |
Projects, Publications, and Open Source Contributions |
|
Evaluating the Memory-Performance Tradeoffs between Hopfield Networks and Standard Attention Mechanisms
[Report] Analyzed performance trade-offs between Hopfield networks and linear attention mechanisms for long-context memory tasks. |
|
Implementing Nested Tensors for Transformer Models
[Report] Reimplemented LLaMA to conserve memory using nested tensors within the attention mechanisms. |
|
Automated Red Teaming for LLMs
[Report] Developed an automated red-teaming framework combining two previous methods to identify LLM vulnerabilities. |
|
SafeGuard
Best Overall Generative Application, UC Berkeley AI Hackathon 2024 [Devpost] [Code] Built an LLM-based security package that detects, classifies, and sanitizes prompt injection attacks on LLMs. Won Best Overall Generative Application from AWS and received a prize of $10K. |
|
Jutsu
[Code] A custom programming language implemented in Python and inspired by popular media. Features include a custom tokenizer and parser, dynamically typed variables, basic control flow, and multi-line function definition. |
|
Augmenting Active Preference-Based Learning of Reward Functions
CS 285 Final Project, 2023 [Report] Developed custom environments and reinforcement learning algorithms for improved intelligent query generation in active learning. |
|
FastMRI
Third Place Grand Award, Intel ISEF 2019 [Code] Reduced acquisition time for MRI imaging using undersampled k-space and a GAN with a custom architecture. Oral Presenter at the National JSHS Symposium. |
|
Publications / Open Source Contributions Visit my Google Scholar profile.
|
Teaching |
|
CS 188: Introduction to Artificial Intelligence
Head Teaching Assistant, Spring 2024 (Cameron Allen, Michael Cohen) Head Teaching Assistant, Fall 2023 (Igor Mordatch, Peyrin Kao) Teaching Assistant, Spring 2023 (Stuart Russell, Peyrin Kao) Teaching Assistant, Fall 2022 (Igor Mordatch, Peyrin Kao) Teaching Assistant, Summer 2022 (Yanlai Yang, Angela Liu) |
|
CS 61C: Great Ideas in Computer Architecture
Academic Intern, Spring 2022 (Connor (Cece) McMahon, Nicholas Weaver) |
|
Other Projects Any awards won are noted in parantheses.
|
|
|