Following one query through PageDB's parser, operators, catalog, table heap, buffer pool, slotted pages, and disk format—and being explicit about the missing transaction and recovery layers.
10 min readdatabases
Unsigned images, a trusted but slow release, destructive database recovery, and a broken ingress path: the experiments that forced Aegis to distinguish configuration from evidence.
11 min readreliability
A gRPC deadline can expire after the provider commits. Why CommerceCore preserves UNKNOWN, retries with durable identity, and reconciles by asking what happened instead of authorizing again.
8 min readcorrectness
The hard part of remote CI is deciding who still owns a job. ForgeCI uses fenced leases so an expired runner cannot publish logs, artifacts, or a terminal result after ownership changes.
7 min readdistributed systems
A leader is only the beginning: persistent consensus state, ReadIndex, request deduplication, snapshots, membership changes, and crash testing made QuorumKV useful.
8 min readconsensus
Primary loss, ambiguous writes, network partitions, synchronous commit trade-offs, and a DELETE that replicated everywhere: the PgSentry experiments that changed how I think about database reliability.
12 min readdatabases
A handshake and one HTTP response were only the checkpoint. WireStack still needed retransmission, RTT estimation, congestion recovery, reassembly, SACK, FIN, and zero-window behavior.
8 min readnetworking
TensorForge's equations expose assumptions; measured workloads expose where those assumptions stop. The model becomes context for regression evidence, not a replacement for it.
7 min readperformance
I started the summer thinking about whether code worked. I ended it thinking about how it fails, how it's deployed, how it's observed, and how it recovers. A retrospective tying together an internship, an event-driven platform, a data lakehouse, and a kernel build.
11 min read
retrospective
Multi-service CI/CD pipelines, security gates that can actually block a merge, HashiCorp Vault, Istio mTLS and AuthorizationPolicy, Kafka and Redis: an honest breakdown of a Summer 2026 DevSecOps internship, written without exposing anything I shouldn't.
13 min read
devsecops
Kafka, Redis, idempotent event ingestion, dead-letter handling, and OpenTelemetry tracing across five services. What building a real event-driven pipeline taught me about failure being the default case, not the edge case.
15 min read
distributed systems
Bronze/Silver/Gold on Apache Iceberg, dbt-on-Trino, Airflow orchestration, idempotent ingestion, and why I removed a favorable-looking performance number in favor of an honest benchmark methodology.
14 min read
data engineering
Building Linux 6.10 from source, writing a character-device driver, and attaching GDB to a paused kernel at start_kernel, and the GCC/C23, LLVM/LTO, and QEMU CPU-model incompatibilities that broke almost every step first.
16 min read
linux
A personal, from-scratch DevSecOps lab: nine VMs, four services in three languages, Jenkins security gates, HashiCorp Vault, Kafka, Istio mTLS, zero-trust AuthorizationPolicy, and Blue-Green deployments. A separate project from any employer's infrastructure.
16 min read
devsecops lab
Lay-Off-Link is an end-to-end MLOps and data platform, not a README fantasy. This breakdown covers Docker Compose through Kubernetes, Helm, Terraform, and a seven-job GitHub Actions pipeline: linting, tests, coverage, multi-image builds, Trivy, k6, and what production-shaped engineering actually taught me.
14 min read
devops
We didn't win. But somewhere between hour 14 and hour 21, I think we found something better. This is the story of how we built Talent Bridge: a web app born from a 24-hour hackathon that reframes layoffs not as career endings, but as transitions.
7 min read
hackathon
\n
Archive / Personal
Earlier writing preserved as a record of what I was learning and thinking at the time.
After spending several months diving deep into Go, I've discovered what makes this language truly special for building reliable, scalable applications. This isn't just another programming language: it's a philosophy that resonates with how I think about software development.
5 min read
go
When I first arrived at the University of Minnesota, I knew I wanted to work in tech, but I wasn't sure exactly where. After exploring different areas and building projects, I discovered DevOps and Site Reliability Engineering: the perfect intersection of systems thinking, automation, and problem-solving. This plan maps out how I'm using my CS and Math double major to build the foundation for a career in infrastructure engineering.
12 min read
learning
My first year at a foreign country, stayed far away from my family, girlfriend, friends, and all memorable things, has come to the end. Now, at the time I wrote this blog, I'm sitting at my desk and typing each line for you to know more about my person.
10 min read
learning
As a computer science student, maintaining a healthy body is just as important as keeping my mind sharp. Over the past year, I've developed and refined this 4-day workout routine that fits perfectly with my busy academic schedule.
8 min read
fitness