About Noel Biju
Detailed background, engineering journey, principles, technology stack dependencies, and development environment.
Personal Story & Journey
My interest in engineering began with a simple question: how do systems behave when disconnected from cloud assumptions? Rather than building standard CRUD applications for textbook assignments, I started taking apart backend architectures, local LLMs, and hardware microcontrollers.
From building my first pharmacy management system (RyMeds) that won Best S1 Project to architecting a multi-tenant laboratory platform with PostgreSQL Row-Level Security (ETLab+), I learned early on that solid database design and security primitives matter far more than visual polish alone.
Today, my focus centers on offline-first AI execution (Ollama, quantized LLaMA/Mistral models), robotics telemetry filtering, and systems programming on Linux.
What I Enjoy Building
Engineering Timeline
Engineering Principles
Core beliefs forged through shipping real applications, debugging sensors, and managing production database schemas.
Simple code usually wins
Simple code is easier to debug, easier to maintain, and easier to explain to someone joining the project later.
When building RyMeds POS counter billing, I chose direct SQL queries over heavy ORM abstractions. This made transaction isolation behavior transparent and prevented implicit database locks.
Building is easier than maintaining
Shipping the first prototype takes days, but maintaining state consistency, handling edge cases, and supporting dependencies takes months.
In ETLab+, adding real-time lab attendance updates required writing clean unsubscription cleanups; without them, long sessions leaked memory on the browser client.
Every bug teaches something deep
System failures reveal invalid assumptions about memory, network boundaries, timing, or concurrency that documentation alone rarely highlights.
Spent hours diagnosing MPU-6050 sensor jitter during my robotics internship, which led me to implement software complementary filters to reconcile low-pass accelerometer data with high-pass gyroscope readings.
Documentation saves future engineers
Writing down architectural decisions, schemas, and trade-offs prevents context loss and saves hours when revisiting codebases months later.
Documenting PostgreSQL Row-Level Security policies in ETLab+'s repository enabled rapid testing when modifying tenant permission schemas.
Small improvements compound quickly
Refactoring one component, cleaning one database query, and writing one unit test per day transforms a fragile codebase into a resilient platform.
Incrementally tuning Ollama model quantization parameters day by day pushed local CPU inference latency from ~2.5 tokens/sec to ~8.5 tokens/sec.
Read error messages completely
The exact line number, call stack, and root failure reason are almost always present in the un-truncated trace if you take time to read it.
Extracted exact kernel boot failure UUIDs from Arch Linux emergency rescue shells instead of guessing hardware driver flags.
Sleep on hard technical problems
Stepping away from the keyboard allows subconscious pattern recognition to resolve complex bugs far faster than brute-force midnight sessions.
Discovered the solution to D3 container dynamic resize re-rendering glitches on mobile charts the morning after a stuck late-night debugging session.
The best tool is the one you understand
Reaching for familiar, transparent primitives yields reliable results faster than introducing heavy frameworks with opaque abstraction layers.
Selected low-overhead FastAPI REST routes paired with SQLite for local memory storage over complex cloud microservice frameworks.
Interactive Technology Stack
Click any technology chip below to filter relevant projects and journal articles using it.
Current Setup & Learning
Active Learning Modules
- ›Quantized GGUF model memory layout & CPU cache optimization
- ›ROS2 (Robot Operating System) node pub/sub communications
- ›PostgreSQL Row-Level Security & index tuning for multi-tenant isolation