I’m a Computer Engineering master’s student at Columbia University with hands-on experience in machine learning, data science and software development. My work bridges high-performance systems and AI, combining C++, Python and SQL with a strong foundation in algorithms, distributed systems and cloud computing (AWS, GCP).
I’ve developed scalable, data-driven applications with PyTorch, scikit-learn, PostgreSQL, Django and Vue.js. My research at Purdue University explored adversarial attacks on Vision Transformers, deepening my understanding of AI robustness and cybersecurity.
I’m passionate about building systems that connect machine learning with software engineering, from optimizing trading platforms to analyzing large-scale spatiotemporal data.
Selected Projects
🏛️ MIT Policy Hackathon 2024 — 1st Place, MIT College of Computing
Collaborated on a policy and technology proposal for a sustainable resentencing system in California. Our solution integrated an interactive public dashboard, machine learning model for case eligibility prediction, and a transparency-based accountability framework for district attorneys.
I led the data analysis and machine learning component, building a Balanced Random Forest classifier trained on ~90,000 cases using SMOTE for class balancing. The model achieved 0.86 recall on eligible cases — effectively identifying candidates for review.
Tools: Python, scikit-learn, SMOTE, Random Forest, Data Ethics
📄 Read Project Report |🧠 Data-Free Backdoor Injection on Vision Transformers (ViTs)
Conducted research on data-free backdoor attacks against the EVA-02 Vision Transformer using LoRA fine-tuning. Injected a stealthy visual trigger without access to the original training data, achieving 84.1% clean accuracy and 60.3% attack success rate (ASR). Implemented GradRollout visualizations to analyze attention-layer activation and proposed a lightweight masking defense to reduce ASR. Tools: PyTorch, Vision Transformers, LoRA, TinyImageNet
Tools: PyTorch, Vision Transformers, LoRA, TinyImageNet
📄 Project Pitch🌡️ EY Open Science AI & Data Challenge — Urban Heat Island Modeling
Ranked 3rd in Latin America (15th globally). Developed a predictive ML model using Sentinel-2, Landsat, and building footprint data; stacked ExtraTrees + XGBoost + ElasticNetCV for optimal generalization and recall.
Tools: Python, scikit-learn, XGBoost, LightGBM, Remote Sensing
📄 Report | 🔗 GitHub🕵️♀️ Spatio-Temporal Crime Patterns Visualization
Built a topological graph analysis framework to uncover crime evolution in São Paulo (2006–2017). Extracted 1,440 subgraphs via TDA and clustered patterns using UMAP + DBSCAN to reveal persistent hotspots.
Tools: Python, OSMnx, TDA, UMAP, DBSCAN
📄 Report🤖 Aurora — Virtual University Counselor
Mental-health chatbot for students using OpenAI, Auth0, and AWS. Integrated keyword alerts, recommendation logic, and privacy-focused data management.
Tools: OpenAI API, Python, PostgreSQL, MongoDB, Vue.js, AWS
📄 Pitch Deck🧩 Transactional Blockchain Simulator with Indexed Query Layer
Designed and built a Blockchain simulation in C++ with core data structures (BST, Hash Table, Linked List). Implemented SHA-256 hashing and Proof-of-Work to ensure tamper detection and historical integrity.
Integrated multiple indexing structures to optimize transactional lookups (range, equality, min/max) over large datasets.
Tools: C++, Cryptography, Data Structures
📄 Report🎨 Cross-Domain Image Colorization with GANs
Implemented GAN-based image colorization (cartoon → real). Benchmarked UNet and ResNet34 architectures using SSIM and CIEDE2000 metrics for objective evaluation.
Tools: PyTorch, GANs, UNet, ResNet34
📄 Paper🔐 RSA & ElGamal Cryptosystem Implementation and Security Analysis
Implemented RSA and ElGamal systems simulating secure Alice/Bob/Eve communication; demonstrated vulnerabilities and applied defensive coding strategies.
Tools: Python, Cryptography, Unit Testing
📄 Report🎮 Virtual Reality Escape Room (UX/HCI)
Developed an immersive VR escape experience with puzzle logic and responsive interactions. User testing achieved >4/5 satisfaction across UX metrics.
Tools: Unity, C#, Oculus SDKs, UX Design
🎥 Demo📷 Multimedia Search & Retrieval System (Image Similarity)
Built a full-stack image retrieval system using Faiss, KD-Tree, and RTree indexing. Vue.js front-end enables image upload and retrieval via Python API.
Tools: Python, Vue.js, Faiss, KD-Tree, RTree
🔗 GitHub