01 / 04

Languages

The foundations I build with

02 / 04

Machine Learning

Where I focus most of my energy

03 / 04

Web Development

Bringing ideas to the browser

04 / 04

Tools & DevOps

What keeps everything running

Multi-agent AI incident response platform that autonomously detects, triages, investigates, and remediates production incidents — with automated post-mortem generation and runbook execution.

Deep learning system that estimates spline-based optical transfer functions to transform medical images between sharp and blurred CT reconstruction kernels using frequency-domain processing with FFT operations.

From-scratch implementation of a 124M parameter GPT-2 trained on 10 billion tokens from FineWeb-Edu, implementing multi-head attention, layer normalization, and autoregressive text generation at scale.

Multi-task CNN-RNN model that classifies paintings by artistic style and artist using a ResNet50-BiLSTM architecture trained on the WikiArt dataset — combining visual features with sequential learning.

AI-powered resume screening platform that lets recruiters search candidates using natural language. Uses FastAPI, OpenAI embeddings, Pinecone vector search, and Supabase for storage.

From-scratch PyTorch implementation of SigLIP — a Vision Transformer pretrained on CC3M with sigmoid contrastive loss and a BERT text encoder for multimodal representation learning.