some stuff i've made
from systems-heavy builds to backend, cloud, and devops work.
Engramsystems · ai infra · backendA distributed memory backend for LLM agents: short-term messages, long-term vector search, pinned facts, auto-extracted knowledge graphs, and cross-session shared memory, all replicated across a Raft cluster.Raft-replicated cluster with LanceDB ANN searchSonarsystems · blockchain · securityLets Solana apps prove a heavy calculation was done correctly off-chain, without redoing it.MVP: historical avg proof in <200k CU, full devnet topologyShadow-Indexsystems · backend · blockchain · databaseReads and stores Ethereum's live activity the moment it happens, so apps can search it instantly.71k+ tx/sec throughputarbxsystems · backend · blockchainA bot that spots price gaps between crypto exchanges and trades them automatically for profit, using borrowed money it pays back in the same instant.17ns AMM calculation, 467us full market scanLithossystems · databaseA from-scratch database storage engine that saves data fast and never loses it, even on a crash.72k+ writes/sec & 97k+ reads/secMempool-Snipersystems · backend · blockchainWatches Ethereum's queue of pending transactions and flags interesting ones in real time.<500ms Detection LatencyMerkle-Trie-RSsystems · blockchainA from-scratch build of the data structure Ethereum uses to prove what's in its database.O(1) State Verificationrusty-redissystems · backend · databaseA from-scratch clone of Redis, a super-fast in-memory store apps use as a cache.~1.4M+ ops/sec (Benchmarked)Aethersystems · database · backendA database spread across many machines that stays correct and online even if some of them fail.Zero-downtime snapshots via VACUUM INTOCrisisMeshsystems · networking · securityA messaging network that keeps working with no internet. phones relay messages directly, encrypted.Guaranteed delivery during partitionsOrionsystems · cloud · devopsA from-scratch tool that runs and heals fleets of app containers across many servers automatically.Self-healing cluster in <2sPrismsystems · database · devtoolsGives every developer their own instant copy of the database to test on, created in a fraction of a second.Database branching in <300msDistributed Task Schedulersystems · backend · cloud · devopsRuns scheduled jobs reliably across many servers, splitting the work and surviving failures.Leader-elected multi-node scheduling