Software Engineer I at Exasol · Database Internals · High-Performance Systems
C++20 • Storage Engines • Cache Dynamics • Mathematical Foundations
"We are born to explore, so explore as much as possible as life is short"
I am a Software Engineer I at Exasol, where I dive deep into database internals, high-performance in-memory analytics, storage engines, and memory hierarchy optimization. I am obsessed with building resilient, deterministic systems that don't break under high concurrency and intense throughput demands.
My engineering approach is anchored in a rigorous appreciation for mathematics—leveraging calculus for rate dynamics and optimization, probability & statistics to analyze tail-latency profiles (p95/p99/p99.9) and thread contention distributions, and linear algebra for multidimensional transformations and data layout design. Whether analyzing CPU cacheline behavior, minimizing lock contention across cores, or implementing Write-Ahead Logs with zero-allocation checksum verification, I prioritize mechanical sympathy and architectural predictability.
Beyond technical depths, I live by the philosophy that "We are born to explore, so explore as much as possible as life is short"—a mindset guiding both my intellectual curiosity across operating systems, compilers, and hardware architectures, and my passion for real-world exploration.
⚡ Featured Systems Project: KV-store
A high-throughput, in-memory Key-Value store engineered from scratch in modern C++20 with zero external dependencies (standard library and POSIX APIs only). Designed for predictable sub-millisecond latencies, thread scalability, and deterministic crash recovery.
| Subsystem | Architectural Design & Implementation | Verified Metric |
|---|---|---|
| Storage Architecture | 32 LRU shards with alignas(64) cacheline padding to eliminate false sharing across CPU cores; doubly-linked list with |
32 Shards (alignas 64) |
| Concurrency & Eviction | Deadlock-free global eviction acquiring all 32 shard locks strictly in ascending index order [0..31]
|
Deadlock-Free [0..31] |
| Write-Ahead Log (WAL) | Binary framing [2B magic "WL", 8B ts, 1B op, 4B klen, 4B vlen, payload, 4B CRC32] with slicing-by-8 checksum computation; configurable sync (always, batch, buffered); deterministic recovery over torn writes |
CRC32 Slicing-by-8 |
| Network & Protocol | Multi-threaded POSIX TCP server natively supporting Redis RESP2 wire protocol (redis-cli compatible) alongside inline commands |
RESP2 Wire-Ready |
| Latency Profile | Sub-millisecond latency under 100k requests (80% GET / 20% SET): p50 = 0.041 ms • p95 = 0.089 ms • p99 = 0.143 ms • p99.9 = 0.312 ms | ~80,451 ops/sec |
| Verification Rigor | 11 C++ unit test suites (115 tests passing: engine, WAL, protocol, stress concurrency, torn writes) + 4-tier Python harness (280 E2E tests passing) | 395 Tests Passing |
Before focusing deeply on database internals and low-level systems programming, college-era open-source contributions served as vital foundational stepping stones. Navigating diverse architectures, collaborating across global developer teams, and merging 257+ Pull Requests cultivated a disciplined instinct for code quality, backward compatibility, and developer empathy.
| Organization / Project | Domain & Focus | Key Contribution Highlights |
|---|---|---|
| Daytona | Developer Infrastructure | Implemented package manager integrations (npm/pnpm tooling switch) in developer environment onboarding workflows |
| Appwrite | Backend Server Core | Enhanced user management core API validation and refined error reporting semantics |
| Golem Cloud | Durable Computing | Engineered Model Context Protocol (MCP) server components and developer tooling integrations |
| InsForge | AI Application Infrastructure | Built cron job monitoring interfaces and responsive UI dialog subsystems |
| Infisical | Secret Management Infrastructure | Implemented Key Management Service (KMS) features within the official Node SDK |



