SynapseX @ GT • Leading 2025
Brain-computer interface organization at Georgia Tech
Role
Front-end Team Lead
Timeline
Sep 2025 – Present
Team
SynapseX @ Georgia Tech
Software team of ~19
Front-end group of 4
Skills
React
Next.js
WebSockets
Python
FastAPI
Overview
Georgia Tech's student-run brain-computer interface org.
SynapseX was founded in 2024 and runs a real research pipeline across neuroscience, signal processing, hardware, and software. The flagship project moves a cursor with your eyes: four targets on screen flicker at different rates, an EEG headset picks up which one your brain has locked onto, and the cursor moves that direction. I led the front-end group on that project and built the org's site.
Research
Neuroscience and signal processing decide what to record and how to clean it, before any of it reaches a model.
Hardware
Device engineering runs the headset: electrode placement, channel count, and getting a usable signal off someone's scalp.
Software
Around 19 engineers in four groups covering the ML pipeline, the dashboard, live streaming, and calibration, on twelve-week cycles with weekly milestones.
Cursor control
Moving a cursor by looking at it.
Four boxes sit at the edges of the screen and flicker at 8, 10, 12, and 15 Hz. Staring at one drives your visual cortex at that same rate, so the classifier only has to work out which frequency is strongest to know which direction you picked. An OpenBCI headset streams the EEG into a FastAPI service that pulls out those frequency features and decodes them.
Pipeline
Operator dashboard
Status for the backend, the socket, the EEG device, and the stream source. Direction arrows light up with each prediction and the cursor carries live coordinates, so whoever is running the demo can see what the headset is actually reporting.
Cursor math on the server
The position started in the browser and drifted between screens. Moving it into the FastAPI service and broadcasting over WebSocket means the operator view and the participant view always show the same cursor.
A gate on the packet path
Classifier packets are ignored until streaming is explicitly turned on, so a stray process cannot move the cursor mid-setup. That is the failure you only find live in front of an audience.
Honest hardware state
The dashboard probes the board before offering live mode. If the headset is not on the port it says so and falls back to a synthetic stream instead of pretending to have signal.
Website
gtsynapsex.com
I designed and built the org's public site in Next.js. It has to explain BCI to a first-year deciding whether to join, and show sponsors what the teams are actually building.