About the project Inspiration:
The growing digital footprint of individuals, coupled with the lack of estate planning, inspired us to build Posthuman. With most people unprepared for managing their digital and financial legacies after death, we saw an opportunity to create a secure, AI-powered solution that preserves memories, ensures financial continuity, and simplifies the transition for loved ones. Too often these days, uncanceled subscriptions and unfound bank accounts take a toll on already grieving family members. We hope that our solution covers use cases for everybody in daily life, and hopefully can be adopted by many.
What it does:
Posthuman is a comprehensive digital estate and legacy management platform that:
- Provides Will execution and asset delegation using Agentic AI and all operated with instructional validity on the blockchain.
- Manages subscriptions and financial accounts post-mortem: (Finding subscriptions from data sources and shifting through and delegating assets)
- Preserves personal interactions through an AI-powered chatbot trained on user data(drives, photos, messages etc)
- Enables scheduled messages, deepfake video recordings, and voice replication for loved ones on occasions etc
- AI Agent Executor executing certain post death interests
How we built it:
We leveraged:
- Mistral 7B for AI-powered interactions and digital legacy recreation, and Ollama to run Mistral
- Elastic Vector Databases to store and retrieve user data locally and Docker for local containers
- LumaAI, ElevenLabs, and Musetalk for AI-based video generation
- EigenLayer for secure blockchain-based AVS verification and execution.
- Infura and Hard Hat for blockchain integration
- Scrapybara to automate post-mortem subscription cancellations.
- OCR and KNN models to analyze financial statements and identify key assets.
- Vercel V0 and Next.js for platform and coding
Challenges we ran into
- Balancing privacy and security: Ensuring all sensitive data remains encrypted and locally stored was a difficult question, as some technologies had to use APIs to achieve our desired outcome/performance
- Realistic AI interactions: Fine-tuning chatbot models to accurately replicate personal conversations was definitely difficult. Sometimes the models were not always perfectly context-aware and had issues providing structured output at first.
- Integrating all the different technologies together was a challenge, and we spent many hour deciding what to use in order to generate the most impactful and useful toolset.
What we learned The importance of privacy-first AI development when handling personal data. How to effectively use vector databases and blockchain for estate planning. So many other new technologies and processes to build software!
What's next for Posthuman
- Introducing multi-user access for family members and executors.
- Integration with additional APIs to gather data more securely and accurately (e.g. Plaid for finance)
- Beta testing with the first few users and potentially partnering with estate planning services and financial institutions for broader adoption
Built With
- docker
- eigenlayer
- elastic
- elevenlabs
- hardhat
- infura
- luma
- mistral
- musetalk
- next.js
- ollama
- scrapybara
- vercel
Log in or sign up for Devpost to join the conversation.