Moaz Mahmud
I design and ship high-throughput backend systems and the data pipelines, retrieval layers, and tooling around them — currently at Enosis Solutions, previously at REVE Systems.
I'm a backend-leaning full-stack engineer based in Dhaka, currently a Senior Software Engineer at Enosis Solutions, where I work on engineering-simulation data platforms, similarity search over legal case data, and LLM-assisted test authoring — mostly in Java, Spring Boot, TypeScript and Python.
Before that, at REVE Systems, I built and led delivery on government-scale systems — office management tools for Bangladesh's Parliament, a bond-license automation platform for 10,000+ users, and a notification system moving roughly a million messages a day.
I grew up moving around Bangladesh — Chittagong, Dhaka, Jashore — following my father's postings, before studying Computer Science and Engineering at BUET. That early exposure to different environments still shows up in how I work: comfortable across unfamiliar codebases, teams, and problem domains.
Outside of production work, I'm a long-time competitive-math and olympiad person — a founding member of Trivuj Math Club and an academic member of the Bangladesh Mathematical Olympiad Committee — and I teach, having recorded 70+ hours of Bangla-language CS and math lectures on YouTube.
- Cut CSV import time by 95% (40 min → 2 min) by streaming and parsing with backpressure and batch writes.
- Stabilized MongoDB writes by eliminating latency spikes via a row-level ingestion buffer.
- Reduced peak load by replacing a heavy cron job with an event-driven space-usage service.
- Lowered LLM cost and latency (−35% tokens, +20% embedding speed) by flattening JSON into concise GraphQL text.
- Improved case retrieval accuracy by integrating Hugging Face embeddings in Qdrant via LangChain.
- Cut test-planning time by 70% by auto-crawling the app and using local LLMs to suggest test steps.
- Accelerated Playwright script creation by 90%, generating runnable tests from those steps with local models (DeepSeek-R1, Qwen-Code).
- Expanded adoption by rewriting a legacy COM add-in into a cross-platform web add-in.
- Streamlined add-in development with a reusable React/Node/TypeScript monorepo.
- Led a team of 3 junior engineers, delivering feedback and bug fixes 2 months ahead of deadline.
- Built a Jenkins pipeline for automated deployment, halving deployment time.
- Fixed a connection-leak bug in the internal DB connection framework, eliminating deadlocks.
- Integrated an internal HTML-to-PDF API, cutting PDF size by 30% with improved quality.
- Architected a multi-channel notification platform (email, SMS, in-app) handling ~1M messages/day.
- Built two reusable vanilla-JS components, cutting boilerplate 10%, adopted by 3 teams.
- Modernized 20% of the codebase with Java Streams, improving maintainability.
- Batched fetching in the reporting framework, eliminating N+1 queries and cutting generation time 50%.
- Delivered 10 training sessions, upskilling 200+ non-technical client staff.
Languages
Backend
Frontend
Data & Messaging
AI / LLM
Build, Test & Ops
Hilbert Curve Space-Filler
An interactive space-filling curve renderer — draws successive orders of the Hilbert curve across a 512×512 canvas, and can map any uploaded image through it as an artistic filter.
2D Ray Tracer & Graphics Sandbox
A graphics-programming sandbox built on OpenGL: a two-axis control wheel, a 2D ray tracer, a custom shader, and fractal animations.
Bangladeshi Number Format Util
A small library converting Bangla or English numeric strings into the Crore-Lakh format — with APIs to both format numbers and spell them out in words.
Multi-Select Table
A JSON-driven table component where rows can be selected and moved between two tables, backed by a JS Map for fast lookups of the selected set.
Tic-Tac-Toe AI
An unbeatable Tic-Tac-Toe opponent built with the minimax algorithm, with gameplay rendered in PyGame.
Mancala Player
An AI opponent for Mancala using minimax search with alpha-beta pruning, with game state organized through OOP design.
Let's work together.
Open to select senior backend and full-stack engineering opportunities. The fastest way to reach me is email.