Kent Worcester
Building efficient, low-overhead applications across blockchains, confidential computing, and other constrained execution environments. Interested in technically challenging work involving low-level engineering, resource optimization, and novel platforms.
Valiant
A high-performance DEX on Fogo, a Solana SVM chain. Work spanned the full stack, including the on-chain program, backend services, and various frontend features and fixes. Major contributions included the site’s real-time chat, “trollbox,” and integration of Fogo Sessions, a session primitive being actively developed alongside our work that required tracking the source as the spec evolved.
Screenshots
Liquidity Book
A full rewrite of Trader Joe's Liquidity Book — a concentrated-liquidity AMM that uses discrete price "bins" instead of a continuous curve — from Solidity to Rust. The CosmWasm contracts target Secret Network, and the frontend is built with Leptos, making the stack Rust end-to-end.
Screenshots
Bitcoin Silent Payments
Implementation of BIP-352: Silent Payments, a protocol for sending Bitcoin payments to a reusable static address without revealing the recipient on-chain. Built for a client-side wallet context as two complementary pieces: a pure Dart library for address generation, sending, and receiving (compatible with the BIP-352 test vectors), and a Flutter plugin that uses Rust via FFI for the performance-critical work of scanning the Bitcoin blockchain for received payments.
Secret Rust
Open-source Rust crates for Secret Network, a Cosmos chain that runs smart contracts inside Intel SGX enclaves. secretrs / secret-sdk-proto provides protobuf types, client interfaces, and the encryption utilities needed to communicate with Secret's enclaves. rsecret is a gRPC client with ergonomics tailored to Secret. keplr-sys provides Rust bindings to the Keplr Wallet browser extension, enabling Rust applications compiled to WASM to interact with the wallet from the browser.
Working with AI
Daily use of AI tools, with a preference for provider-agnostic ones like opencode that decouple the workflow from any single model or inference provider. Comfortable running models locally on consumer hardware with llama.cpp, including model quantization, KV cache quantization, MoE expert offloading, and sampler tuning. Also exploring image, video, speech, and music generation.