079
Projects Blog About
[PUBLICATION] ARVO: Atlas of Reproducible Vulnerabilities for Open Source Software
[PUBLICATION] ARVO: Atlas of Reproducible Vulnerabilities for Open Source Software

We introduce ARVO: an Atlas of Reproducible Vulnerabilities in Open-source software. Sourcing vulnerabilities from C/C++ projects that Google’s OSS-Fuzz discovered, we successfully reproduce more than 9,000 memory vulnerabilities across over 250 projects, each with a triggering input, the canonical patch, and the ability to automatically reproduce the project from source.

ResearchLLMSecurity

JavaScript Art
JavaScript Art

Some artworks with JavaScript and creative coding.

JavaScriptWebGL

Monte Carlo Simulation
Monte Carlo Simulation

Conducting monte carlo simulation of a portfolio of stocks for the next 100 days.

Monte CarloCholesky

[PUBLICATION] Comparative study of the ansätze in quantum language models
[PUBLICATION] Comparative study of the ansätze in quantum language models

We evaluate the performance of quantum natural language processing (QNLP) models based on these ansätze at different levels in text classification tasks.

ResearchQuantum ComputingNatural Language Processing

Deep Learning S&P 500
Deep Learning S&P 500

A deep learning classification model that predicts daily direction of the S&P 500 index

Long Short-Term MemoryRecurrent Neural NetworkFeature Engineering

Spread Trading Nasdaq and S&P 500 when JPOW Increases Interest Rates
Spread Trading Nasdaq and S&P 500 when JPOW Increases Interest Rates

We explore how we can take a short position in the spread options on the e-mini Nasdaq-100 and the e-mini S&P 500 index futures. Then, we price a spread option using multiple stochastic factors, using monte carlo simulations.

Trading Strategy

Constructing Social Network using Foundational Language Models
Constructing Social Network using Foundational Language Models

Sneak peek what I learned back at JPMorgan about leveraging Name and Entity Recognition (NER) models to construct a social network graph based on "bond strengths." We use cosine similarities to calculate bond strengths between embedded entities including companies and individuals like Elon Musk, Apple, First Republic Bank, etc.

LLMPython

Punishing Infants
Punishing Infants

Meritorious Award at SIMIODE Challenge Using Differential Equations Modeling Competition

Math ModelingMarkov ChainMatplotlib

C++ Inference Engine
C++ Inference Engine

An inference engine built from scratch with C++. Parses PyTorch model into ONNX model, loads the model and input data, and make fast inference calls using C++ binary.

C++Deep Learning

with by 079
theme portfolYOU