About
01 / SYSTEMS
Every design decision begins with understanding the full system boundary. Complexity is the enemy of reliability, and clever architecture is debt that compounds.
02 / OBSERVABILITY
Assume every dependency will fail. Design for graceful degradation. Invest in observability before you need it, not after an outage.
03 / PERFORMANCE
In Matching Engine, p99 latency is a product requirement. Profiling, GC tuning, and lock-free data structures are first-class engineering concerns.
Background
I am a senior backend engineer with 5+ years of experience specializing in Matching Engine infrastructure, low-latency order matching, and high-reliability systems.
My expertise spans Java, Spring Boot, DevOps, and building robust financial infrastructure. I have a proven track record in developing complex trading systems, orchestrating cloud-native architectures on AWS EKS, and creating internal productivity tools that drive team productivity.
I specialize in sub-millisecond latency order matching systems, distributed backend architecture, and building systems that prioritize correctness and operational reliability under real production load.
My focus is on system performance optimization, reliability engineering, and creating tools that simplify operational complexity. I believe in building systems that are easy to reason about, debug, and operate at scale.
Projects
Production-grade systems, proxies, and observability tools built for modern engineering stacks.
Geospatial Intelligence
Fullscreen interactive 3D geospatial intelligence globe inspired by WorldView-style OSINT dashboards. Real-time visualization of aircraft tracks, satellite orbits, global disaster alerts, weather overlays, and active surveillance CCTV feeds.
LLM Infrastructure
High-performance caching and streaming deduplication proxy for LLMs. Reduces token usage costs on metered APIs by caching identical prompts and coalescing concurrent in-flight requests into a single multicast upstream stream.
Kubernetes Security
Production-safe, auditable, read-only file exploration agent for Kubernetes containers. Solves security and compliance challenges of inspecting container filesystems in production without requiring shell access or elevated permissions.
Microservices Infrastructure
Lightweight Node.js service registry, discovery service, and operational control plane. Supports heartbeat-based registration, weighted round-robin / consistent hashing routing, and Redis lease leadership control plane mechanisms.
Explore additional tools and repositories on my GitHub profile.
Systems Knowledge
Technologies studied to production depth, not tutorial depth.
JVM Platform
Distributed Systems
Messaging & Streaming
Kubernetes
Security & Protocol
Writing
Notes from the trenches. No tutorials. Just hard-won operational understanding.
2024 · 12
The hidden cost of sidecar proxy interception: iptables overhead, per-connection TLS setup time, and measuring the real latency tax of mTLS in high-throughput services.
2024 · 10
How reactive streams backpressure signals propagate through system boundaries, where they silently break, and what actually happens to your Kafka lag when a consumer falls behind.
2024 · 08
Rolling upgrades on StatefulSets behave very differently from Deployments: ordered pod termination, PVC lifecycle edge cases, and the race conditions that bite you during node drains.
2024 · 06
Reading the Raft paper gives you the algorithm. Running it in production teaches you about leader lease expiry, snapshot transfer interruptions, and follower log divergence under write amplification.
2024 · 03
Most GC problems are allocation problems in disguise. Profiling methodology, identifying hot allocation sites with async-profiler, and the surprising places object churn hides in JVM applications.
2024 · 01
Kafka transactions guarantee exactly-once delivery, but at a price. Epoch management, producer fencing, and the latency implications of transactional producers in high-throughput systems.
Current Focus
eBPF — Kernel observability and network policy enforcement without kernel modules or service mesh overhead
Linux Scheduler Internals — CFS, task groups, and the impact of NUMA topology on tail latency
JVM Vector API — SIMD intrinsics through Java's Panama project for numeric compute workloads
Kubernetes Gateway API — The next generation of traffic management beyond Ingress
Deterministic Simulation Testing — FoundationDB-style correctness verification for distributed systems
My current learning is oriented toward the intersection of kernel-level observability and application-level performance. eBPF represents a fundamental shift in how we can instrument production systems without performance overhead or invasive code changes.
I am also deeply interested in deterministic simulation testing, a technique pioneered by FoundationDB where an entire distributed system is simulated deterministically, allowing bugs that would take months to surface in production to be reproduced in seconds.
These two directions represent the same underlying thesis: the future of reliability engineering is a shift from reactive observability toward proactive correctness verification.
Resume