-
Trust and Teach AI DApp - On-chain RLHF
Decentralized Reinforcement Learning with Human Feedback (RLHF) DApp running on-chain using Cartesi rollups. Enables trusted, specialized, and aligned LLMs.... Read more
Tags: RLHF, LLM, Blockchain, Ethereum, Cartesi, Decentralized AI
Date: 11/10/2024
-
Synth Dev - LLM Documentation Understanding
Helping LLMs understand developer documentation by creating diversified use case coding examples for pre-training and RAG. Tools: Restack, TogetherAI, Llama, LlamaIndex.... Read more
Tags: LLMs, Documentation, Python, AI, Data Generation, RAG
Date: 11/9/2024
-
Personal Portfolio Website
A modern, brutalist-inspired personal website built with Astro, showcasing projects, blog posts, and personal information.... Read more
Skills: Web Development, Frontend, Static Site Generation, TypeScript
Tags: Astro, TypeScript, Web Development, Static Site
Date: 2/19/2024
-
Lightlink Hackathon - AI Data Peer Review
A Web3 peer-review system for creating high-quality AI datasets, focusing on teaching AI how to think. Built for the Lightlink Hackathon. [Solidity, Web3, API3 QRNG, Blockchain]... Read more
Tags: Web3, Blockchain, AI, Solidity, Smart Contracts, Peer Review, Lightlink, Hackathon
Date: 2/4/2024
-
TeachAI [hackathon]
Our application helps to curate high quality datasets by providing a framework to incentivize RLHF, with the LLM being fully verifiable and hosted on-chain. https://ethglobal.com/showcase/teachai-pkr1z 🥇 Best New Subgraph/Substream 🥇 Cartesi — Best MVP I put LLM into L2 on-chain using cartesi (linux L2). user to LLM interaction through onchain making the data transparent and LLM deterministic and trusted this is the way to do trusted ai... Read more
Tags: Large Language Models (LLM), Web3, Linux, Solidity
Date: 10/1/2023
-
Fund AI Training DAO [hackathon]
Our new type of DAO allows us to crowdsource and manage the multi-stage process of creating large AI models. We showcase the value of crowdfunded AI with its ability to find bugs in solidity contracts. https://ethglobal.com/showcase/fund-ai-training-dao-2isnr WINNER OF 🚀 XDC Foundation — Crowdfounding 💎 DELV — Best UI Integration 🥉 XMTP — Best Use Powerful AI models should be a public good, just like Linux, Ethereum, and academic research on AI... Read more
Tags: Large Language Models (LLM), Decentralized Autonomous Organizations (DAO), React.js, Solidity
Date: 8/1/2023
-
IRL-trust [hackathon]
The on-chain interpersonal credit score for unbanked https://app.buidlbox.io/projects/irl-trust Prizes: from HUMA: A credit line for every wallet from Ceramics: best identity application Key innovation: credit score based on common connections. In contrast to traditional credit score that is for each account, interpersonal credit score is unique fore every 2 people rather than per person IRL-Trust https://github.com/kirilligum/irl-trust The on-chain interpersonal credit score for unbanked... Read more
Tags: Fraud Prevention, React.js, Credit, Solidity
Date: 2/1/2023
-
Sw/.tch - Decentralized Professional Network
Decentralized professional network for engineers with JSON resume on IPFS and smart contract for referral tokenomics. Built with FEVM, IPFS, Solidity, TypeScript.... Read more
Tags: Web3, Decentralized Network, Resume, IPFS, FEVM, Solidity, TypeScript
Date: 11/20/2022
-
ZK KYC SBT [hackathon]
zero-knowledge know-your-customer and personhood proof soulbound token https://ethglobal.com/showcase/zk-kyc-sbt-ozwb2 WINNER OF 🚀 Optimism — Just Deploy! 🥇 WalletConnect — Best DeFi 🏊♂️ Worldcoin — Pool Prize First-ever noncustodial privacy-preserving legal compliance for onboarding and defi.... Read more
Tags: React.js, Solidity
Date: 7/1/2022
-
Command Line Tool for System Statistics
A cross-platform CLI tool that provides detailed system performance and resource utilization statistics.... Read more
Skills: Python, System Performance, CLI Development
Tags: Python, System Monitoring, Performance, CLI
Date: 6/15/2022
-
CompliSend [hackathon]
aml payments - KYC erc20/721 transfer with messaging and zk reporting to an auditor https://ethglobal.com/showcase/complisend-3j0jx WINNER OF 🏊♂️ XMTP — Pool Prize 🥉 Mina — Best Use transfer only if the recipient has KYC_SBT (non-transferable NFT with zk proof of identity that we built at eth NYC)... Read more
Tags: React.js, Solidity
Date: 4/1/2022
-
Sensor Suit for Exercise Detection
A suit with 12 motion sensors connected using SPI forwards data to smartphone, where algorithms distinguish between different types of activities.... Read more
Skills: Embedded Systems, C++, Machine Learning
Tags: Arduino, AVR, feature extraction, anomaly detection
Date: 2/1/2015
-
Modified extremely randomized trees
Cache-optimized implementation of extremely randomized trees. [C++, cache-optimization, cache-oblivious]... Read more
Tags: C++, cache-optimization, cache-oblivious
Date: 10/1/2014
-
Predicting insurance re loss (Kaggle competition)
Given 450K labeled samples with 300 features each, predict insurance loss on another 450K samples. [Python, pandas, sklearn, unbalanced data, missing values, feature importance, mixture of models] Using cross-validation on R^2, gini index and F1-score, I compared a number of combinations of procedures for features selections (Lasso, ensemble trees, PCA, SVM), data cleaning/transformations (one-hot, ordered/unordered, unbalanced), imputing missing values (mean, multiple imputation, GMM, ensemble methods), regression models (gradient boosted and extremely randomized trees), and stacking of different predictive models using averaging and non-linear regressions.... Read more
Tags: Python, pandas, sklearn, unbalanced data, missing values, feature importance, mixture of models
Date: 8/1/2014
-
Multi-factor trading
= Data preparation ---- custom cvs parser ---- filtering data ---- checking for errors = Trading algorithm ---- identifying features ---- implementing random forest and other machine learning regression and classification algorithms = Testing ---- comparisons between strategies Tools: C++, Boost, OpenCV... Read more
Tags: C++, Boost, OpenCV
Date: 3/1/2014
-
Educational games recommendation platform
Founded and co-lead a team of 8 people at a startup mixer to follow lean startup approach. Customer discovery with over 10 hypothesis-validation interviews a week. Established partnership with content providers and schools. Developed a WebApp prototype. Associated with UC Berkeley.... Read more
Date: 3/1/2013