1/ Over the last few months, we've been working to build one of the first credit scores for DeFi.
Today, we're excited to share the results of our first credit score with the world! π
Read more below π
Julian Gay
5,884 posts
- Hopeful for a better future for SFUSD. Excited to vote in this special election cc @recallsfboe
- Replying to @juliangay2/ Quick information about us: Our mission is to bring DeFi lending to 1 billion people. We believe that DeFi will one day provide access to financial resources to anyone with an internet connection. Key to this? Quantification of risk at scale. π
- Replying to @juliangay5/ Our machine-learning based credit score is initially built on historical borrowing activity from the @aave v2 liquidity protocol. Inputs include behaviour and other time-based account attributes enabling us to predict propensity for liquidation in the near future.
- Replying to @juliangay3/ One of the main reasons quantification of risk is so important is that, without it, DeFi lending has to remain heavily over-collateralized = inaccessible. The next 1B users of DeFi will have real-world use cases that will require more capital-efficient forms of lending.
- We went through @defialliance accelerator last year which connected us with top crypto investors. Looking to accelerate your biz and build a strong network? Last day to apply to next batch *Jan 7*:
- Replying to @juliangay and @AAve6/ The dataset powering this model is built from the health factors and collateral / debt composition of all Aave v2 accounts at 15 minute intervals from their first Aave interaction until present. It contains over 360M observations. β¨
- Replying to @juliangay4/ More economic activity moving on-chain will also lead to new, diverse applications where accurate, transparent risk scoring will personalize and improve the experiences that are possible for consumers. Helping enable this is our small part to play.
- Replying to @juliangay and @AAve7/ Health Factor (HF) is the core statistic to maintaining Aave protocol solvency and removing bad debt (i.e. under-collateralized debt) from the system. It is defined as:
- @AliMCollins talks about "working effectively as a team" while actively suing 5 out of 6 other Commissioners and #SFUSD. Farcical! cc @recallsfboeCommissioner Alison Collins, who was stripped of her VP title on the SF school board after posting racist tweets against Asian Americans, on why she doesn't support Commissioner Moliga as the board's new VP. Ultimately, the board voted 6-1 to elect Moliga. Collins voted no.
00:00 - Replying to @juliangay and @AAve15/ If you'd like to view your own score, you can do so by signing up to our Cred Score Beta API waitlist here:
- Replying to @juliangay @AAve and @cred_protocol19/ Thank you to @AaveGrants, @AaveAave, @stani and our ever supportive investors for their help with our journey so far.
- Replying to @juliangay and @AAve9/ The high level dataset design is shown in the image below. We plan to release a post explaining how we developed this high-fidelity dataset soon.
- Replying to @juliangay and @AAve11/ Some of the features included in this classifier were: β Account age π Aggregations of the time series of historical HF π Interactions with the Aave protocol π΄ The types of assets an account borrows and keeps as collateral






