Welcome to My Neurotechnology Journey

After years building real-time AI systems and graphics engines, I'm making a deliberate transition into neurotechnology and brain-computer interfaces. This blog will document my journey—from foundational coursework to hands-on BCI projects.

Why Neurotechnology?

After reaching a point where I wanted to use my abilities to create lasting impact rather than just commercial success, I discovered the groundbreaking work at Science Corp—restoring vision to the blind through cortical implants. Combined with my fascination for the human brain, I realized neurotechnology represents the perfect convergence of meaningful impact and technical challenge.

I'm particularly interested in:

Closed-loop neural implants for treatment-resistant depression
Communication systems for locked-in patients
Cortical visual prosthetics like Science Corp's pioneering work
Computational neuroscience of cognition and decision-making

What I'm Learning

My preparation involves systematic skill-building across multiple domains:

Computational Neuroscience

Understanding how neurons encode and decode information, from spike trains to population coding and neural dynamics.

Signal Processing

Real-time filtering, artifact removal, and feature extraction from noisy EEG/LFP data—essential for robust BCI systems.

Machine Learning for Neural Data

Classification algorithms optimized for the unique characteristics of neural signals, including motor imagery and P300 detection.

Example: Basic EEG Preprocessing

Here's how you might filter motor imagery EEG data in Python using MNE:

import mne
from scipy import signal

def bandpass_filter(raw, l_freq=8, h_freq=30):
    """
    Apply bandpass filter for motor imagery (mu/beta bands)

    Parameters:
    - raw: MNE Raw object
    - l_freq: Lower frequency bound (Hz)
    - h_freq: Upper frequency bound (Hz)
    """
    return raw.filter(l_freq, h_freq, fir_design='firwin')

def apply_car(raw):
    """Remove common noise across channels"""
    return raw.set_eeg_reference('average', projection=True)

def preprocess_motor_imagery(raw):
    """Complete preprocessing pipeline"""
    # Bandpass filter
    raw_filtered = bandpass_filter(raw)

    # Common Average Reference
    raw_clean = apply_car(raw_filtered)

    return raw_clean

Mathematical Foundations

The Information Transfer Rate (ITR) for a BCI system measures how much information can be communicated per unit time:

ITR = \log_2(N) + P \log_2(P) + (1-P) \log_2\left(\frac{1-P}{N-1}\right) \times \frac{60}{T}

Where:

$N$ = number of possible selections
$P$ = accuracy (0-1)
$T$ = time per selection (seconds)

For example, a P300 speller with:

36 characters ( $N = 36$ )
85% accuracy ( $P = 0.85$ )
5 seconds per selection ( $T = 5$ )

Achieves: ITR ≈ 17 bits/minute

What's Coming Next

Follow along as I build and document:

Motor Imagery BCI Classifier - EEG-based control using Common Spatial Patterns
P300 Speller System - Communication interface for assistive technology
Real-time Neural Decoder - Kalman filters for movement prediction
Signal Quality Assessment - Automated EEG data quality tools

Each project will have detailed technical posts explaining the neuroscience background, algorithmic approaches, and engineering challenges encountered.

The Goal

Long-term, I want to work on clinical-grade brain-computer interfaces that make real differences in patients' lives—whether that's restoring communication, treating depression, or giving vision back to the blind.

Let's build something meaningful together.

→ Get in touch if you're working on neurotechnology or have insights to share.