Skills

Python / PyTorch

Deep experience building custom training loops, autograd hooks, and model surgery for interpretability experiments.

Computer Vision

Building and deploying vision models — object detection (YOLOv3), multi-view CNNs for medical imaging, and real-time inference pipelines.

Mechanistic Interpretability

Research into reverse-engineering neural network circuits to understand internal representations — induction heads, feature superposition, and causal patching.

Applied Math (PDEs, Numerical Methods)

Foundational work in partial differential equations, variational modeling, and numerical analysis — bridging physics and machine learning.

Frontend Engineering (React, Mapping/Geo)

React/Next.js frontends with interactive data visualization, geographic mapping, and real-time annotation interfaces.

Projects

The Decompiled Mind

An interactive interpretability experiment: a small transformer trained from scratch on a synthetic induction task, hand-ported to run live in the browser, with validated attention visualizations and causal patching.

Lung Nodule Detection — LUNA16

A multi-view CNN (three ResNet-18 branches over axial/coronal/sagittal CT slices, fused for prediction) trained on the public LUNA16 dataset to flag candidate nodules for review — detection only; diagnosis stays with clinicians.

Automated Exam Monitoring

A real-time proctoring prototype using YOLOv3 and OpenCV to detect unauthorized objects and multiple people in frame, with inference testing under Linux.

RIFT-NYC

Bridges mechanical engineering and ML: pipe-stress analysis and PDE/variational modeling applied to a scalable backend architecture, with physics-informed neural network experimentation tracked via TensorBoard.

Experience

Web Developer — Early Ruskin Manuscripts Archive

Own the frontend for a digital archive of 19th-century John Ruskin manuscripts: interactive zoom/annotation viewer for handwritten primary sources, backend integration (Nginx, Express, PHP, TEI XML), and geographic mapping tying manuscripts to locations. Built documentation enabling non-technical staff to manage the archive independently.

Senior Mathematics Tutor — SELU

Promoted from Tutor to Senior Tutor; led sessions in Calculus I-III, Linear Algebra, and Differential Equations; mentored junior tutors; built structured problem-solving frameworks.

Computer Vision & Data Science Intern — Nawaratna EBS

Built an online exam monitoring prototype (YOLOv3/OpenCV), converted legacy student records to structured databases, created performance dashboards with Pandas/Matplotlib.

See exactly how this is generated →