← Home

Writing

Notes on backend engineering and security research — what broke, what I shipped, and what I'd do differently.

Latest · System design

Clinic booking system design: PostgreSQL locks and ADRs at MAL 4

How our three-person team designed a clinic appointment prototype at MAL 4 using PostgreSQL row locks, a partial unique index, idempotency keys and 19 ADRs.

02 Nov 2026 · 9 min read

All posts

Infra

Self-hosting college infrastructure with Coolify, Mailcow and Discord

How I run ACM MPSTME's self-hosted infrastructure: Coolify deployments, Mailcow email for 7–9 committees, member mailboxes and Discord-based access.

· 9 min read
Research

AI vulnerability variant analysis: VARSEER benchmark results and limits

My VARSEER study compares clone matching, retrieval signals and LLM triage on 49 vulnerability families across 12 C projects, including negative results and uncertainty.

· 16 min read
Backend

Maintaining a student portal: booking rules, PostgreSQL and admin workflows

How I maintain MPSTME's railway-concession portal for approximately 4,000 students: server-side booking rules, PostgreSQL history, admin tools and daily email reports.

· 12 min read
Backend

Building a QR check-in system: PostgreSQL race conditions and role-based access

How I built Social Conclave's QR check-in system for 500+ delegates, with editable role permissions, manual scanning fallback and PostgreSQL duplicate prevention.

· 4 min read
Applied crypto

End-to-end encrypted file sharing: GhostShare's browser encryption design

How I designed GhostShare's encrypted notes and file sharing with browser-side Argon2id, AES-256-GCM, direct uploads to Cloudflare R2 and explicit trust limits.

· 5 min read
Systems

Building a database storage engine in Java: B-Trees, tombstones and compaction

How I built bad-db, a Java storage engine with a binary record format, an in-memory B-Tree index, tombstone deletes and compaction, and what it lacks for durability.

· 5 min read
Security

LLM prompt injection: lessons from building a jailbreak contest

What running Fool The LLM taught me about prompt injection, encoded secret leaks, streaming-output checks and authorization outside the language model.

· 5 min read