Inspiration
Traditional computing systems operate on a highly rigid, explicit input model. When a user experiences intense cognitive overload or a workflow blockage, they are forced to seek out documentation or summon help menus manually. This requirement introduces significant operational latency, fractures focus, and heightens cognitive friction. Overall, there's a 40% drop in accuracy and a 112% drop in reliability.
I was inspired to build CGEN-NeuroSync to transform software from a passive tool into an active, cognitively aware model. Human flow, thought, and user interfaces should work hand-in-hand; therefore, I built a system that detects focus blockages and clears them before the user even realizes their productivity is dropping.
What it does
CGEN-NeuroSync is a localized, adaptive system architecture designed for closed-loop cognitive state modulation. Using raw, passive electroencephalography (EEG) telemetry, the system continuously monitors a user's neurological state without requiring any active physical inputs.
When a cognitive overload or frustration spike is identified, the backend instantly triggers an interface adjustment, dynamically morphing the layout or generating assistive UI assets to seamlessly resolve the workflow blockage.
How we built it
The core system is built on a dual-engine machine learning backend consisting of two proprietary AI models: Agent Synapse and Agent CGEN.
Data Pipeline: Raw EEG telemetry streams into a localized processing pipeline where custom algorithms filter out muscular artifacts and baseline noise in.
Synapse Model: This model computes user-specific neural baselines and continuously maps changes in frequency bands to determine current cognitive load states.
CGEN Model: Operating downstream, this generative engine converts the classified cognitive state into precise UI modulation parameters, adapting the software workspace programmatically.
The underlying structural framework and execution logic of this closed-loop architecture have been formally documented and are currently patent-pending under an active Provisional Patent Application (PPA).
Challenges we ran into
The primary engineering hurdle revolved around signal corruption from muscular artifacts (such as eye blinks, facial movements, and jaw clenching). These movements generate massive voltage spikes around the frontol cortex that mask authentic neurological activity:$$V_{total}(t) = V_{neural}(t) + V_{artifact}(t)$$Because $$V_{\text{artifact}}(t) \gg V_{\text{neural}}(t)$$ standard threshold filters often experience high false-positive rates. We overcame this by designing an adaptive, localized filtering matrix capable of isolating and subtracting artifact noise without introducing systemic processing latency. Additionally, we had to solve the issue of individual neural variability, as universal frequency thresholds fail across different users. I resolved this by building an automated, rolling baseline calibration sequence into the Synapse initialization phase.
Accomplishments that we're proud of
Noise Mitigation: Successfully developed a localized processing node capable of isolating clean neurological signals from raw, noisy biometric data in.
Provisional Patent Protection: Translating a theoretical neuroscience framework into a fully modeled, defensible architecture and successfully securing patent-pending status (PPA).
Closed-Loop Execution: Achieving a truly functional, low-latency loop where raw telemetry directly drives automated interface optimization without manual user prompts.
Validated localized signal processing system against standard open-source MNE/PhysioNet human EEG datasets, demonstrating mitigation of muscular artifacts and active tracking of dynamic cognitive load shifts.
What we learned
Building CGEN-NeuroSync enhanced my understanding of digital signal processing (DSP), multi-modal AI orchestration, and advanced human-computer interaction (HCI). I learned how to manage asynchronous data streams efficiently and gained practical experience navigating the complexities of protecting intellectual property in the deep-tech software space.
What's next for CGEN NeuroSync
Now that the architectural foundation is built and patent-protected, the next phase involves expanding the platform into a highly scalable, commercial B2B SaaS product. I plan to optimize the dual-model backend to integrate seamlessly via cross-platform developer APIs, allowing enterprise software ecosystems to naturally leverage cognitive-aware optimization to maximize workplace efficiency.
Log in or sign up for Devpost to join the conversation.