Systems and backend engineer. I build things that are close to the metal — distributed consensus, GPU compute, bytecode interpreters. Rust-first, correctness-minded, benchmarks required.

Async key-value store in Rust/Tokio with a Redis-compatible wire protocol implemented from scratch (no client library). Supports AOF persistence, pub/sub, TTL eviction, and an in-progress Raft consensus layer for fault-tolerant replication. Extended with a Python MCP server for tool-based access.

The interesting part isn't the storage — it's building the RESP parser and Raft state machine from first principles.

GPU N-body gravitational simulation using wgpu/WGSL compute shaders. Uses tiled shared memory and ping-pong buffering to minimize global memory reads. Criterion-benchmarked against a naive CPU baseline on an RTX 3060 Ti — meaningful speedup at scale.

If you've done GPU work, you know the real challenge is memory access patterns. That's what this is about.

Stack-based bytecode VM in Rust. Defines its own 7-opcode instruction set, compiles to bytecode, and includes a disassembler for inspecting generated instructions. Built to understand how language runtimes actually work under the hood.

Primary: Rust, Tokio, async systems
GPU/compute: wgpu, WGSL
Backend services: Python, FastAPI, PostgreSQL

Open to backend and systems engineering roles — distributed systems, infrastructure, performance engineering. Not looking for frontend or customer-facing work.

salasdeveloper99@gmail.com · github.com/salas-dev-max