Nahom Zewdu

Dark
Dark

Selected Projects

A selection of systems built around technical problems.

A selection of systems I've built to explore backend infrastructure, developer tooling, and knowledge systems. Each project represents solving a technical problem rather than showcasing a technology stack.

Nuvom

A Python task execution engine designed for developers who need reliable background processing without the operational overhead of traditional distributed task queues.

Highlights

  • Plugin-oriented architecture
  • Automatic task discovery
  • Retry and scheduling support
  • Multiple serialization strategies
  • Structured logging and observability
  • Production-oriented execution model

Existing solutions often assume external infrastructure or prioritize distributed deployments over simplicity. Nuvom explores a lightweight architecture focused on portability, predictable execution, and developer experience.

Designed around modular components including pluggable queue backends, result backends, task discovery, scheduling, retries, batching, observability, and configurable execution strategies. The architecture emphasizes extensibility while keeping the developer experience straightforward.

Status: Actively evolving as an open-source infrastructure project.

KMS

An AI-powered knowledge management system that transforms engineering conversations, repositories, and documentation into a searchable knowledge base for teams.

Highlights

  • Multi-source knowledge ingestion
  • AI-powered semantic retrieval
  • Repository and documentation indexing
  • Modular ingestion pipeline
  • Extensible architecture for new data sources

Engineering knowledge is often fragmented across Slack, GitHub, documentation, and individual contributors. KMS focuses on preserving institutional knowledge and making it accessible through semantic search and retrieval.

Built around an ingestion pipeline that processes multiple knowledge sources into a unified retrieval system. The architecture combines document processing, embeddings, vector search, and LLM-powered retrieval while remaining modular enough to support additional integrations.

Status: Under active development.

Guessit

A real-time multiplayer drawing and guessing game focused on synchronization, low-latency communication, and multiplayer state management.

Highlights

  • Real-time multiplayer architecture
  • WebSocket communication
  • Room management
  • Live synchronization
  • Event-driven game engine

Building responsive multiplayer applications requires careful handling of synchronization, game state consistency, and real-time communication between multiple participants.

Designed around WebSocket-based communication with centralized game state management, room coordination, player synchronization, and event-driven interactions to provide a smooth multiplayer experience.

Status: Functional deployment with ongoing improvements.