Portfolio

Parallel Swap for Ref Finance

Participated in hackathon bounty contest to optimize the allocation of token trades amongst parallel automated market maker (AMM) pools. Used Lagrange Multipliers to determine optimal route allocation to maximize output tokens. Implemented for the Ref Finance project on the Near Protocol. Interfaced with Web3 Smart Contracts to measure current state of pools, and then implement optimal swap commands.

Smart Route V2

A generalization of the parallel swap project, used convex optimization and Lagrange multipliers to determine optimal route allocation of input tokens, given combinations of routes with up to two network hops among pools in the AMM network for Ref Finance.

Nutrition Dashboard

An ongoing project in which I am using nutrition data from different staple foods, and different nutrition axes such as vitamins, protein, energy, etc., along with constraints of food preferences to determine optimal minimum set of foods to satisfy the nutrient profile. Desire to make it usable by anyone wanting to grow a self-sustainable home garden and for crop planning. Implemented using a convex optimization library in python and a streamlit frontend.

Chainlink Hackathon 2021: PETA-Bot

My first foray into Web3: a hackathon project where I implemented a Twitter bot and a price edge indicator, along with an analysis dashboard. The goal was to show when a particular decentralized exchange (DEX) had a price edge compared with the market price measured by Chainlink price oracles.

Chainlink Hackathon 2022: Adaptive Market Maker

My second big hackathon project: Adaptive Market Maker (AdMM). In this project, I wrote and deployed my own smart contracts in Solidity on the Ethereum testnet, and built an interacting web front-end. The goal was to establish a proof of principle that you could use volitility data from Chainlink price oracles to determine the optimal "Trading Curve" shape for a given token pair. More stable pairs would have a more linear trading curve, while more volatile pairs would induce a trading curve with much more curvature, tunable by a single parameter q. Based idea off an interesting white paper on contstant power root market makers.

Gitcoin Hackathon: ENS Integration

This was an interesting project where I had to dive into using Clojurescript -- a dialect of Lisp that compiles Javascript code. I also used the Reagent framework and reverse-engineering a hashing function for determining a reverse-ENS lookup. This is basically like a DNS lookup for ethereum addresses that are mapped to a specific name, such as vitalik.eth

Problem Daily

This is a work in progress. The idea is to build a set of physics problem / word problem "mad libs", where we can use a pseudo-random number generator to build a huge number of unique physics problems and solutions for helping students and teachers build intuition for solving physics problems. I had used a Google Cloud Function (a scalable serverless solution similar to Amazon AWS Lambda functions) for implementing my python code for the problem generation, but ran out of the free trial credits. So I recently converted the problem generator google cloud function into an AWS Lambda Function.