Jani

Jani

Inspiration:

Jani was not born in a boardroom, but at home, watching my grandfather's eyes. Seeing his Alzheimer’s episodes worsen every day witnessing how his reality. Becomes a puzzle with missing pieces ignited a fire in me. I built Jani because I wanted to give him an anchor, a voice that never tires of answering his questions, and a tool that helps him hold onto his identity for as long as possible.

What it does:

Jani is a specialized voice assistant designed to be a "Cognitive Anchor" for Alzheimer's patients. It provides:

Continuous Validation.
Answers repetitive questions to reduce "Sundowning" anxiety. - Neuro-Rehabilitation: Uses active stimulation to slow cognitive decline. - Phonetic Adaptation: A custom engine that understands fragmented or soft speech.

How I built it

Built on an Edge-AI architecture to ensure privacy, Jani utilizes:

RAG (Retrieval-Augmented Generation)
To access a personal "Biographic Database."
SRT Algorithm: To schedule memory reinforcements based on the forgetting curve.

Challenges I ran into

- Emotional Resilience: Coding while watching my grandfather's condition worsen was the greatest hurdle. It turned the project into a personal race against time. - The precission: Adapting LLMs to be "Infinite-Patient" and never corrective.

Hardware Optimization: Ensuring the model runs locally with latency<250.

Accomplishments

I'm proud of seeing Jani successfully calm my grandfather during a crisis episode. - Phonetic Inclusivity: Integrating Aymara-based phonetics to bridge the speech gap. - Dignity Restoration: Providing a tool that respects the user as a person, not just a patient.

What I learned

I learned that the rate of synaptic decay can be mitigated by high-frequency, low-stress emotional engagement Technology is only as powerful as the empathy behind its code.

What's next for Jani

- Family: Cloning family voices to increase comfort levels. - Clinical Pilots: Expanding from my home to geriatric centers across the region. - Predictive Analysis: Forecasting agitation states before they escalate.