mohamedrafee2002@gmail.com | Portfolio | LinkedIn

I’m a software engineer focused on distributed systems, data systems and backend infrastructure.

I enjoy working on systems from first principles - understanding how they work at a fundamental level and building them with a focus on correctness, reliability and clear trade-offs. My work spans from low-level primitives like consensus and storage engines to higher-level distributed architectures.

I also write about system design and engineering, sharing lessons from building real-world systems.

I primarily work on systems and backend engineering:

• Languages: Java, TypeScript, Python, Rust - building backend services and system components
• Systems engineering: Concurrency, Distributed systems and storage systems
• System design: Microservices, event-driven architectures, coordination and consistency models
• Technologies : PostgreSQL, Elasticsearch, ClickHouse, Redis, Kubernetes, Prometheus, Grafana, AWS - Lamda, S3, DynamoDB, SQS etc.

I’m interested in exploring foundational systems and infrastructure by building them end-to-end - implementing and understanding the underlying mechanisms from scratch to go beyond abstractions.

I’m building this to understand how real distributed systems and coordination work - combining consensus, storage, failure modes and system design.

Axis is a fault-tolerant, strongly consistent distributed key-value store designed for coordination, distributed locking and metadata management - inspired by etcd-io/etcd.

• Raft-based replication with leader election, failover and crash-safe recovery
• Multi-Versioned storage engine (MVCC) with ACID semantics enabling non-blocking temporal queries
• Linearizable reads
• Leasing system with checkpointing

I built this to deeply understand how consensus works in real systems - leader election, replication and maintaining consistency under failures - inspired by etcd-io/raft.

Jaft is an implementation of the Raft consensus protocol in Java, built as the consensus layer for Axis.

• Pure state machine design with determisitic execution model
• Leader election, log replication and joint consensus for dynamic cluster membership
• Snapshot via log compaction, Quorum confirmed and lease based linearizable reads, Leadership transfer

I built this to understand and learn storage engines and MVCC works

An embeddable multi-versioned storage engine enabling time-travel queries built on LMDB.

• Implements a non-blocking single-writer, multi-reader concurrency model and timeline of historical states
• Uses versioned-index, an ordered key-value index based on Persistent (copy-on-write) B+tree for key timline indexing enabling efficient lock-free reads and atomic writes
• Supports snapshot isolated point and range reads, batched writes and incremental background compaction

Thanks for reading - always up for a conversation about systems, trade-offs, and interesting engineering problems.