Github

Mezzanine Labs Internship Projects

BillingModule Solidity Contract Implementation

This contract integrated the overall Mezzanine contract structure as a feature of a department or treasury, allowing each department in a company to track its billing on-chain. I devised this abstracted contract for each department to decrease query time when creating financial statements, from O(num departments x num bills) to O(num bills). As more users are onboarded, this is an estimated O(n^2) to O(n) improvement. This feature supported the searchability and improved query speed of financial transaction data on the blockchain.

Mezzanine Contract Test Suite

I wrote a majority of the smart contract test suite using Foundry. I was able to catch three critical contract inheritance vulnerabilities before sending our contracts for auditing, saving us a lot of costs. I wrote simple assertTrue tests, and more complicated fuzz tests and invariant testing in more than 50% of the contracts. Since many of the contracts depended on instances of each other, often the most complicated part was the setup function for testing high-level functionality such as the BillingModule, which needed me to deploy a large portion of company contracts (e.g. MezzHub, Fee controller) to simulate the intended billing usage (in a department).

Backend Subgraph

Mezzanine needed a way of interfacing the frontend UI with the blockchain state needs to be developed. I created a backend subgraph to index blockchain events to facilitate efficient frontend interaction with the blockchain, which has handled over 10,000 requests. I designed the GraphQL schema based on the existing contract dependency tree, created blockchain event mappings, and deployed the subgraph on The Graph’s indexers. Later, Mezzanine needed our own data indexers, so I used Terraform to configure a small-scale AWS EC2 instance to host a graph-node indexer for our custom blockchain. After implementing the subgraph, I integrated the GraphQL API with the ViteJS front end and created a comprehensive testing suite in TypeScript. To

Mezzanine Frontend Onboarding Application

We received multiple user onboarding requests, and each session took approximately three or more hours. I automated the 180-minute onboarding process to take less than 20 minutes. I also worked on designing and implementing a payment flow. First, I used Figma to prototype the front end. After the proposal was approved, I created a ViteJS + TailwindCSS implementation of the flow. I used Stripe, RainbowSDK for payments, and AirtableJS (later PostgreSQL) for storing user onboarding data.

Personal Projects

xv6-riscv OS Demand Paging (DP) & Copy-on-Write (COW)

https://www.loom.com/share/48725d5f658f430ebeb2c262a3022df3

https://www.loom.com/share/dc63dadf34a849959a4ff6e1c3c3133f

I implemented both DP and COW memory management in the xv6-riscv OS, reducing memory use by 61%***** in a sample workload while maintaining performance. This project required precise modifications to the kernel's memory subsystem, particularly in files vm.c, trap.c, kalloc.c, kalloc.h and pa_track.h

• Extended virtual memory system with custom page table entry flags as well as modification of physical memory allocation and data copying in xv6 to support deferred allocation
• Implemented a synchronized reference counting system to track shared pages across processes - preventing memory leaks
• Created trap handlers for DP and COW page faults with proper TLB coherence management
• Specific kernel-to-user space data transfer modifications to handle demand allocated pages and copy on write pages.
• Identified and mitigated race conditions in concurrent memory access scenarios and process scheduling (sbrkfail in usertests.c)
• Verified implementation using original xv6 usertests and created performance tests demonstrating memory savings and trade-offs

Here’s a demonstration of the tradeoffs: