Inspiration
Echo & Ancestor (under the Kukhanya vision) was born from the realization that indigenous African heritage is facing "Silent Atrophy." As our elders pass away, their oral traditions—the primary vessel for our history—often vanish with them. I was inspired to bridge the gap between tech-native youth and their ancestral roots by moving beyond static archives and creating an interactive mentorship experience that preserves the emotional presence and rhythmic wisdom of a digital elder.
What it does
Kukhanya is a sophisticated "Memory Twin"—a living repository that combines high-fidelity vocal presence with a culturally-grounded reasoning engine. Using the Gemini Multimodal Live API, it allows users to have real-time, interruptible voice conversations with an AI modeled after ancestral mentors. It doesn't just retrieve facts; it interprets the world through indigenous frameworks, such as the "River of Time," and can even "see" cultural artifacts via camera to narrate their historical and spiritual significance.
How we built it
The system’s foundation is built on Google Cloud infrastructure, utilizing Gemini 1.5 Pro for its massive context window. I have developed the initial RAG (Retrieval-Augmented Generation) pipeline and established a professional repository and folder architecture to house our growing knowledge base of transcripts and cultural data. For the interface, I utilized Lovable to prototype a 'Campfire Sanctuary' UI, and I am currently orchestrating the backend via Vertex AI to manage real-time multimodal interactions and ensure the system remains grounded in our primary source materials.
Challenges we ran into
One of the primary technical hurdles was implementing "Ontological Guardrails." I had to move beyond standard LLM outputs to ensure the AI stayed faithful to indigenous metaphors, preventing it from defaulting to Western industrial logic. Additionally, transitioning from a turn-based chat model to a truly "Live" multimodal agent required significant optimization of the latency between visual recognition and vocal response to maintain an "authentic presence."
Accomplishments that we're proud of
I am incredibly proud of my conceptual framework for Semantic Filters and Source-Grounded Verification. By hard-coding linguistic reverence for sacred identities, I have successfully moved beyond the 'chatbot' paradigm. I have demonstrated that AI can be a intentional vessel for cultural safety, ensuring that personal and collective histories are treated as living legacies rather than just generic data points for retrieval.
What we learned
Building this project taught us that the future of AI isn't just about speed or intelligence, but about contextual resonance. I learned how to leverage Gemini’s native interleaved output to weave together narration with visuals seamlessly. I also gained a deep understanding of the ethics involved in "Digital Immortality" and the importance of preserving the "vocal texture" of a culture to keep its spirit alive.
What's next for KUKHANYA
The architecture for Kukhanya is designed as a universal vessel for heritage, though its heart remains personal. While our immediate priority is perfecting the experience for my own family and community to ensure our stories are preserved with dignity, the underlying system is built to scale. If there is a high demand for this type of intimate, secure digital preservation, we intend to make the platform available to the world, allowing other lineages—from Maori to Native American—to build their own interactive bridges to the past.
Built With
- app-engine
- cloud-run
- cursor
- gemini-1.5-pro
- gemini-multimodel-live-api
- github
- google-cloud
- loveable
- n8n
- python
- rag
- react
- vertex-ai
- vetcor-search
Log in or sign up for Devpost to join the conversation.