Stephen.dev
Back to Portfolio
Architectural Deep-DiveSystems Engineer

Mental Health Diary Backend

A ultra-high-performance, low-memory footprint server architecture built to process high-concurrency requests. Implements a customized engine processing binary audio data stream to execute instant automated text transcriptions via asynchronous endpoints.

Rust (Actix-web)PythonPLpgSQLPostgreSQL

System Challenge

Processing continuous, high-concurrency binary audio streams for transcription while maintaining an ultra-low memory footprint and strict privacy standards for sensitive mental health logs.

Architecture & Solution

Architected the core ingestion service in Rust using Actix-web for bare-metal performance and memory safety. Offloaded heavy transcription machine-learning tasks to a Python microservice, bridging the two via efficient PLpgSQL functions in PostgreSQL.

Performance Outcomes

  • Achieved a 90% reduction in RAM overhead compared to a Node.js equivalent.
  • Implemented robust asynchronous queues capable of handling simultaneous audio stream decoding.
  • Secured all payloads with robust in-transit and at-rest cryptography.