Case studies that document the problem, the decisions and trade-offs, and what actually got built — across backend, DevOps, hardware and local AI.
12 projects shown
A local-first learning and engineering workspace: ingest study material, turn it into a visual DAG roadmap, and prove progress through competency gates — with FastAPI, PostgreSQL and a Gemini → Ollama → offline AI cascade.
Ultra-fast Bangladeshi Phone Number, Geo District, and SME Customer Normalizer with E.164 standardization and Bengali numeral support.
MikroTik RouterOS & ISP Uptime Watchdog with Sub-Second Jitter Detection and Telegram Outage Alerts.
A resilient background orchestration plane managing asynchronous local tasks, health checks, cron triggers, and bidirectional Telegram alerts.
A local-first document processing pipeline that extracts handwritten and printed Bengali khata ledger entries into structured Excel and Google Sheets without cloud API costs.
An experimental, fully local AI orchestration layer designed to manage personal tasks, learning workflows, and system automation without cloud dependencies.
A lightweight, headless automation daemon that sends scheduled due payment reminders and invoice receipts to retail pharmacy clients with zero paid API overhead.
Benchmarking prompt evaluation times, token generation speeds, and VRAM quantization for local models on Ollama.
High-availability router configuration, firewall filtering rules, VLAN segmenting, and 24/7 uptime surveillance benches.
Configuring headless GPU rendering pipelines, custom resolutions, and low-latency virtual display buffers on Linux.
Engineering custom battery management systems (BMS), multi-transistor audio amplifiers, and resilient DC solar power setups.
Architecting a multi-platform bridge embedding high-throughput web dashboards into native Android runtimes with offline caching.