Ember AI

Ember AI

Crypto as easy as texting a friend

Created on 1st July 2024

Ember AI

Ember AI

Crypto as easy as texting a friend

The problem Ember AI solves

Problem:
Achieving a single goal in DeFi can require visiting multiple websites, signing even more transactions, and clicking for minutes on end, plus waiting for transactions to settle before the next one can be started, and DeFi UX is majorly held back by GUI constraints. Current intent-based approaches to DeFi, even if addressing the need for a one-click-checkout type of experience using cross-chain liquidity, still require dApp teams to build custom "recipes" for integration, and to deploy those to their front-end before end users can reap the UX benefits.

Solution:
We’re tying together the full intent pipeline. From extracting user intents from natural language, to aggregating asset routing and intent-filling as it exists today, to dApp integration and “recipe” composition.
Just talk to Ember about what you want to do in DeFi, and how you'll pay. Ember sets it up for you, and you sign once using your existing self-custody solution. No browsing around, no confusing language, no needless waiting.
Message Ember on Telegram or our Farcaster Frame.
Our Widget allows dApp teams to integrate into Ember by answering 5 simple questions and adding 1 code snippet to their frontend code. This widget brings the full Ember experience to their website, saving thousands in front-end development costs, and enabling never-before possible customer service experiences and behavioral analytics.
This stands in opposition to the current low-access paradigm of great DeFi UX. User goals are born in natural language, and we’re removing all complexity and extra steps between that and an outcome.
Ember currently conducts various cross-chain transactions, and we're integrating the best protocols all over DeFi, to allow users to interact with them and compose DeFi primitives through intent-based systems using natural language.

Demo video is slightly outdated. Now feature signing with existing self-custody wallet

Challenges I ran into

  • Our first solution for intent classification was too inaccurate so we researched and experimented with new methods until performance was optimized.
  • ⁠Farcaster Frames and the various frameworks seem quite immature so we had to navigate the build despite poor documentation.
  • ⁠LLM AI agent frameworks were extremely opinionated adding way too much latency to response times. We learned how to minimize LLM inferences to drastically improve the user experience.
  • Classification of DeFi primitives into a valid and useful structure for the Action Graph was a fun bump to work through.

Tracks Applied (1)

Payments Track

We're building crypto for your mom, re-imagining how users interact with their funds on and off chain, through a convers...Read More

PAYMENTS with Stripe

Cheer Project

Cheering for a project means supporting a project you like with as little as 0.0025 ETH. Right now, you can Cheer using ETH on Arbitrum, Optimism and Base.

Discussion

Builders also viewed

See more projects on Devfolio