Muhammed Shafin P

Posted on Dec 3

My Journey Creating Qeltrix: A 17-Year-Old's Approach to Cryptographic Innovation

#hejhdiss #opensource #qeltrix #python

Why I'm Sharing This

I'm sharing my journey and knowledge to inspire others - especially tech-loving and open-source-loving people. This is how I learned, how I built, and how I think. Maybe it can help or inspire someone else.

About Open Source: Right now, my mindset is that I like to donate my works to open source. I believe in sharing knowledge and code freely with the community. I don't know what the future holds or how my thinking might change, but today, this is what I believe in - contributing to the open-source ecosystem and making my work available for everyone.

Who Am I and What Did I Build?

I'm 17 years old, and I created Qeltrix (V1-V6) - a cryptographic file container system. But here's what makes this journey unique: I didn't write most of the code traditionally. Instead, I designed the entire architecture and leveraged AI as a tool to bring my vision to life.

My Role: Architect, Not Traditional Developer

What I Actually Did

System Architecture & Design

Designed the complete Qeltrix architecture from scratch
Made all technical decisions: what algorithms to use, how components interact, system flow
Researched extensively to choose the right technologies and approaches
Designed the parallel processing model, streaming architecture, and cryptographic pipeline

Research & Technology Selection

Researched cryptographic libraries and algorithms
Studied parallel processing techniques (SIMD, multi-threading, process pools)
Investigated compression strategies and entropy analysis
Compared different technical approaches before making design decisions

AI-Assisted Development

Created detailed prompts for Gemini to generate code
Generated most code using AI (primarily Gemini)
I edited, reviewed, and validated generated code
Verified cryptographic implementations were correct
Ensured code matched my architectural vision
Used Claude and Gemini for creating documentation and explanations

Quality Assurance

Checked all Python code thoroughly despite limited coding experience
Validated cryptographic correctness and entropy outputs
Tested different modes and configurations
Verified the system worked as designed

CRITICAL NOTE - Please Verify Yourself: My quality checks may be wrong. I may have missed errors or made incorrect validations since I'm not a professional developer. The quality assurance was done to the best of my ability with my current knowledge level, but you should not blindly trust my validations. If you're using Qeltrix, please check and verify the code yourself. My validations are not professional-grade.

My Knowledge Foundation: Self-Taught and Curiosity-Driven

Important Context About My Knowledge

What's listed here are just samples - not my complete knowledge base. I've learned many more things (SIMT, various other concepts and technologies) that aren't listed here.

About "Forgotten" Knowledge: I may have forgotten specific details of things I learned, but they're not truly gone. If I read about them again or happen to hear about them, they come back to memory quickly or just want read again. I don't need to learn them hardcore again - the foundation is there.

Depth vs Coverage: What I'm sharing here may not be deeply mastered professional-level knowledge. Instead, I know the outer coverage - I understand how things work, how components connect, and the fundamental concepts. This is why you shouldn't see me as a professional developer. I'm different - I'm a curiosity-driven learner who understands systems broadly rather than specializing deeply.

My Self-Assessment: Honestly, I think I'm still at a very basic level - especially when compared to others my age. I don't have a strong foundation in anything specific - my knowledge is broad but not deep. I understand concepts and how things connect, but I'm not an expert in any single domain.

The JEE Reality: Now I need to focus on JEE preparation. The irony? JEE tests PCM (Physics, Chemistry, Mathematics) - not Computer Science. I don't know if I'll get into college because the exam doesn't test what I'm actually passionate about and knowledgeable in.

About Mathematics in JEE: While I have strong logical and analytical thinking from CS, JEE Mathematics is a different beast - it requires extensive practice in calculus, algebra, trigonometry, coordinate geometry, and more. It's not just about understanding concepts; it's about speed and accuracy under exam pressure. This is where I need to focus my efforts now.

But I need to pass it to pursue engineering education.

How I Actually Learn

My learning isn't traditional. I don't follow structured courses - I follow curiosity. When something interests me, I dive deep until I understand it. When I see something I don't know, I research it immediately.

My Test Lab: My own computers and VMs. Everything I learn, I test hands-on.

What I Know (Theory-Strong, Practice-Learning)

Programming & Languages

Python basics (variables, functions, control flow)
Assembly language concepts (registers, memory management, CPU operations)
Bytecode understanding (learned after discovering JIT compilation)
Explored multiple programming languages for curiosity - even if forgotten, the concepts remain valuable
Smali concept (understood it's the human-readable bytecode format for DVM and ART, learned while studying AhMyth)
Understanding of compilers: AOT (ART) vs JIT

Computer Science Fundamentals

SIMD (Single Instruction Multiple Data) operations
Parallel processing concepts
CPU architecture basics from assembly learning
Display systems: DRM (Direct Rendering Manager), driver layers, kernel graphics stack
Wayland vs Xorg architecture and compositor systems
Understanding of different managers: display managers, window managers, session managers

Cryptography (Theory)

Understanding of how cryptographic algorithms work conceptually
Entropy, randomness, and security principles
Key derivation, encryption/decryption processes

Networking & Security

TCP protocol fundamentals
OSI Model (especially Layer 2, and understanding of other layers)
ARP protocol: opcode 1 (request) and opcode 2 (reply)
Scapy: Can use for ARP sniffing and sending packets
Ethical hacking concepts and methodologies
BPF (Berkeley Packet Filter) Concept - learned while studying usage of Wireshark

Finance & Trading

Crypto trading concepts and technical analysis
Learned technical aspects of cryptocurrency and trading mechanisms

My Main Interest Domain: Tech

Artificial Intelligence and machine learning concepts
Emerging technologies and their applications
How different technologies integrate and interact
Broad tech ecosystem understanding

Systems & Linux

Multi-boot setups using Ventoy
Linux system architecture and systemd targets (multi-user.target vs graphical.target)
Default CLI interfaces and switching between display servers
Driver layers and kernel module interactions

My Hardware Reality: Learning on a 2018 HP Laptop

My Test Machine: HP bs0xx (bought 2018, my dad's laptop)

Intel Core i3-6006U (dual-core, 2.0 GHz)
4GB RAM
Very low-spec by today's standards

What I Did With It:

Installed Debian (chose it for low resource usage)
Configured default boot to multi-user.target (CLI only)
Initially installed Xorg, learned it was resource-heavy
Researched display server architecture
Learned about DRM, driver layers, kernel graphics stack
Switched from Xorg to Wayland for better performance
Optimized every component for this low-spec machine

Why This Matters: When you're running on 4GB RAM from 2018, you MUST understand what every component does. This forced me to learn system architecture deeply.

Real-World Learning Examples

Example 1: The Display Server Journey

Installed Debian with default Xorg
System felt slow → researched why
Learned: Xorg → DRM → kernel drivers → GPU
Discovered Wayland as alternative
Switched to Wayland, understood the entire graphics stack in the process

Example 2: The Bytecode Curiosity Chain

Saw mention of JIT compilation somewhere online
Got curious → researched bytecode concepts
Friend mentioned AhMyth (ethical hacking tool)
Tried to read its source code → written in Smali
Learned Smali is the human-readable bytecode format for Dalvik VM (DVM) and ART
Understood the concept: DVM → ART compiler → AOT vs JIT
This knowledge later helped me understand BPF in Wireshark
Understood how Scapy works at a deeper level

Example 3: The Qeltrix Testing Process

Tested every Qeltrix version on another laptop.(Lenovo loq)
Validated cryptographic outputs manually
Benchmarked on real hardware (the published 44.8 MB/s results)

My Learning Style: Curiosity → Research → Test → Understand

I don't learn linearly. I learn through curiosity chains:

See something interesting or unknown
Research immediately until I understand it
Test it hands-on on my own systems
Connect it to other knowledge
The concept becomes part of my knowledge base

Even if I forget specific details later, the fundamental concepts remain valuable and help me understand new things faster.

My Interest Distribution

High Interest & Deep Knowledge: Computer Science, Networking, Security, Systems
Medium Interest: Physics and Chemistry (can grasp interesting concepts easily, but less passionate)

What Makes My Knowledge Different

Not a Traditional Coder, But...

I understand systems architecture deeply
I know how components interact at multiple layers
I can validate implementations even without writing them
I test everything hands-on on my own hardware
I research obsessively when curious
I connect knowledge from different domains

My Qeltrix Work Wasn't Just V1 PoC

Designed the entire architecture from scratch (V1 through V6)
Researched every component and technology choice
Tested on my own low-spec hardware
Validated cryptographic correctness
Created all documentation and explanations
Managed the entire project vision and direction

My Knowledge Is Practical, Not Just Theoretical

Every concept tested on my own machines (laptop + VMs)
Actual packet crafting with Scapy, not just reading
Hands-on cryptography validation, not just theory

What I Don't Know (Yet)

Advanced Python libraries for cryptography implementation
Professional-level coding practices and optimization
How to write complex cryptographic code from scratch in Python
Production-level software engineering
Physics and Chemistry at the depth I know CS

The Reality of AI-Assisted Development

This project demonstrates a new paradigm: you don't need to be an expert coder to architect complex systems.

What AI Handled

Writing advanced Python library usage (cryptography, multiprocessing)
Implementing complex algorithms based on my specifications
Handling syntax and library-specific details
Generating boilerplate and structure

What AI Couldn't Do Without Me

Architecture decisions: AI can't design system architecture
Research and selection: AI can't choose the right approach
Quality validation: AI can't verify its own correctness in context
Vision and direction: AI can't set project goals and requirements
Cryptographic verification: AI can't guarantee security properties

The Reality of AI Output vs My Vision

Important Truth: The final code output is approximately equal to my vision, but not entirely accurate.

No matter how clearly I explained things to the AI, it makes its own differences and interpretations. This means:

I had to constantly review and validate outputs
Some implementations deviated from my exact architectural vision
I edited and corrected where the AI misunderstood
The final result is a collaboration, not a perfect translation of my design

This is the reality of AI-assisted development - it's a powerful tool, but still requires constant human oversight and correction.

My Validation Process

Even though I didn't write the advanced Python code:

I understood what every function should do architecturally
I verified the cryptographic implementations matched theory
I tested outputs (entropy, compression, correctness)
I edited code when it didn't match my design
I understood memory flow and parallel processing implementation

Key Point: You can validate code quality through testing, understanding theory, and architectural knowledge - even without being able to write it yourself from scratch.

Learning Through Building

This project became a massive learning experience:

Cryptography: From theoretical knowledge to practical implementation validation
System Design: First major architecture project
Parallel Processing: Applied theoretical knowledge to real implementation
AI Collaboration: Learned how to effectively use AI as a development tool

The Hard Truth: Time vs. Reality

I'm now pausing this project completely. Why? JEE exam preparation.

I'm 17, and I have about 1.5 months until JEE Session 1. I need to get into a good engineering college for B.Tech. This isn't just an exam - it's my entire future career path.

My situation is uncertain. I don't know if I'll get into college or not. Every day of preparation counts. If I continue spending time on Qeltrix now, I risk compromising my already limited chances of college admission, which would directly impact my ability to become an engineer professionally.

JEE must come first. My education and future career depend on it.

What Happens Next?

After JEE: I may or may not return to Qeltrix, depending on:

College workload and commitments
Career direction and opportunities
Other priorities in college life

The project remains fully open-source under GPLv3 (code) and CC BY-SA 4.0 (concepts). Anyone can continue, fork, or build upon this work.

Message to Aspiring Builders

You don't need traditional coding skills to architect systems. Here's what you actually need:

Theoretical understanding of how things work
Research skills to find the right approaches
System thinking to design architecture
Critical validation to verify implementations
AI literacy to effectively use tools like Gemini/Claude

If you're young, curious, and have theoretical knowledge: you can build complex systems. AI bridges the gap between architectural vision and implementation.

My Knowledge Style: Cross-Domain Connections

My knowledge is vast across many different domains - and that's my strength. I don't just learn individual topics in isolation; I connect knowledge across domains.

How I Apply Cross-Domain Knowledge

When I work on one domain, I actively try to apply knowledge from another domain. This creates unique solutions and deeper understanding:

Cryptography concepts → Applied to file container design
Networking knowledge → Applied to understanding parallel processing
Bytecode understanding → Helped understand BPF concepts
System architecture thinking → Applied to designing Qeltrix

This cross-pollination of knowledge from different fields is what makes my approach different.

What I'm Learning (Now Paused for JEE)

Networking / CCNA Topics: Before pausing for JEE, I was learning networking fundamentals and CCNA-level topics to understand networks more deeply.

About AhMyth and How Qeltrix Started

I tried to understand AhMyth (a powerful and dangerous Android RAT used in ethical hacking). I still don't know how to use it - I dropped it. It was just curiosity-driven exploration.

If I remember correctly, I started Qeltrix when I was just searching about Smali and what it is. That curiosity about bytecode formats led me down a path that eventually resulted in designing an entirely different project - a cryptographic file container system.

Acknowledgments

To Gemini: For generating the code that brought my architecture to life.

To the Community: For testing, feedback, and contributions.

To Future Contributors: The architecture is solid. The code works. The concept is proven. If someone wants to take this further - especially in Rust or C++ - the foundation is ready.

Learning Under Financial Constraints

My Financial Reality: I come from a low-income background and learned everything in a resource-constrained environment. I don't have access to expensive hardware, paid courses, or premium tools.

My Hardware Journey:

Started with: 2018 HP bs0xx laptop (my dad's laptop with 4GB RAM, Intel i3-6006U)
This was my main development and testing machine for Qeltrix
My Testing Broke Things: Sometimes I broke my OS while testing and experimenting. I tried so many things, tested so many configurations - breaking and fixing my system was part of my learning process.
Got Lenovo LOQ before going to coaching center
Where I Actually Built Qeltrix: I went to coaching center (it was hostel-based, as you know) with the hope that I would learn for JEE. But I couldn't take my LOQ to coaching center since it's heavier. I can't risk destroying it since it's my device that I'll have forever. So I took that old HP laptop under the name of "learning for JEE" - but even though I hoped I would focus on JEE preparation, my curiosity started taking over there. I ended up creating Qeltrix at the coaching center instead, using that old HP.

Zero-Cost Learning Approach:

Used only free and open-source resources
AI for Learning: Used free tiers of ChatGPT, Gemini, and Claude - not just for code generation, but for LEARNING. I chatted with them to understand concepts and get resources the way I wanted to learn.
- Example: I learned ARP protocol by chatting with ChatGPT
- Used AI to explain concepts in ways that made sense to me
- Got explanations, examples, and resources through AI conversations
Online references, free documentation, YouTube videos, and online classes
All tools and software: completely free (Debian, Python, open-source libraries)
No paid courses, no expensive books, no premium subscriptions
No resources, no connections, no mentors - purely self-taught

Why My Knowledge Might Be Incorrect: I don't have professional guidance or structured education in CS. I created my own syllabus, chose what to learn based on curiosity, and had no one to correct my understanding. I learned through AI chats, online resources, and self-experimentation (even breaking my OS multiple times). This is why my validations and knowledge might be wrong - I'm learning in isolation without expert feedback.

My Curiosity Goes Beyond Tech: I have so many interests and curiosities:

Metallurgy Example: At one point, I was learning about alloys and thought about creating a new alloy at home (homemade). I stopped due to limitations - when I started researching, I realized I'd need dangerous materials like lead(in gas form - metal fumes) and would need to handle temperatures of 800-1000°C+. Not possible in my situation, so I stopped at the start.
Computer Testing: I tested my own computer extensively with different configurations, trying to understand hardware limits and optimization - sometimes breaking my OS in the process and having to fix it
Many other curiosities that I explored until hitting resource or safety limitations

Resource Constraints Forced Deep Learning: Having limited resources actually made me learn more deeply. When you can't just buy better hardware or paid solutions, you MUST understand how things work at a fundamental level to optimize and work within constraints.

The Pressure at Home: I have constant pressure from home to focus on JEE preparation. I acknowledge and thank my parents for this - they're pushing me because they want what's best for my future. They're right to be concerned.

But I can't help it. My interests and curiosity drag me here. Even when I went to hostel-based coaching center hoping I would finally focus on JEE, my curiosity still took over. Even when I know I should be studying Physics, Chemistry, and Math for JEE, my mind wanders to cryptography, system architecture, networking, and building things. It's not a choice - it's who I am.

My financial situation and lack of resources never stopped me from learning - it just changed HOW I learned. Everything I built and learned cost zero money, just time and curiosity driven by passion I can't control.

Final Thoughts

Qeltrix proves something important: architecture and theoretical knowledge matter more than coding ability when building innovative systems. Modern AI tools can bridge implementation gaps if you have:

Clear vision of what you want to build
Understanding of how it should work
Ability to validate correctness
Persistence to iterate until it's right

I'm 17, and I built a working cryptographic system by combining theoretical knowledge with AI assistance. That's the future of development.

Now I need to focus on getting into college so I can become an even better engineer.

Ethics & Important Clarifications

About Ethical Hacking and Security Testing

Please don't take this the wrong way: I have strong ethics and boundaries when it comes to security and hacking.

What I've NEVER done and NEVER will do:

Hack another person's devices or accounts
Test on systems I don't own or have explicit permission to use
Use my knowledge to harm, steal, or invade anyone's privacy
Access unauthorized networks or data

What I've actually tested on (all legally and ethically):

My own home router
My own phone
My own computers and laptops
Virtual machines (VMs) that I created and own
Only devices and networks that belong to me or where I have explicit permission

About Android Security: I'm not an expert in Android security. As I mentioned, I was learning networking concepts - AhMyth was just a curiosity point that led me down the Smali/bytecode learning path. I didn't drop it because I didn't understand it - at that time, I also happened to explore a language called Haskell (which is for mathematics and functional programming, and I also dropped it later). Haskell was also in my learning plan, but Qeltrix started happening before I could really jump into it. My curiosity fluctuates when something newer comes along, but some things remain permanent interests like networking, AI, backend development, and similar domains.

Example of Multi-Domain Knowledge: This ability comes when you have knowledge across multiple domains. For instance, I already know how to use tools like Metasploit and Nmap. Even for tools I haven't directly used, I can understand how they work if they're within my knowledge domain. For example:

Show me a tool and its command-line arguments
I can tell you if it uses UDP or TCP scans
I understand the underlying protocols: TCP, UDP, and others
I know the basic meaning and differences between protocols
I understand the difference between normal Ethernet (802.3) and Ethernet II
I can read a tool's usage and understand what it's doing at the protocol level
I can analyze Wireshark packet captures because I know TCP flags (SYN, ACK, FIN, RST, PSH, URG) and what each flag means
When I see packets in Wireshark, I understand what's happening in the communication because I know the protocols and flags

My Current Networking Learning: I've basically completed learning ARP and was looking into VLAN, and then STP (Spanning Tree Protocol). For testing and experimenting with these, I was planning to use Linux namespaces since I don't have money to buy real networking devices for testing and experimentation.

This is because I understand the fundamentals. When you know how TCP works, how UDP differs from TCP, and how Ethernet frames are structured, you can understand any tool that operates at these layers - even if you've never used it before. You just need to see the command syntax and you can understand what's happening under the hood.

CRITICAL CLARIFICATION ABOUT AI USAGE AND CODING: Let me be absolutely clear - I CAN code. I have always been able to code. Using AI for code generation doesn't mean I can't code. Never think that way.

My Coding Background: I'm sure I can write code - I learned what was taught in Plus Two (NCERT curriculum) and much more that I learned on my own. Maybe I'm still at a basic level when compared to professionals, but I have coding skills. If I didn't have coding skills, I couldn't have edited the AI-created high-level library usage in the code, reviewed it, validated it, and corrected it when needed.

I used AI as a tool to accelerate development and handle advanced library implementations, but I understand programming, I can write code, and I've always had coding ability. The AI was a productivity tool, not a replacement for coding skills I didn't have.

About My Coding Abilities

Let me be clear and honest: When I say "I used AI for code generation," it doesn't mean I can't code at all.

What I CAN do:

Write basic to intermediate Python code
Understand code logic and structure
Debug and fix code issues
Read and comprehend code in multiple languages

What I meant about AI assistance:

I cannot yet write advanced or professional-level code from scratch
Complex cryptographic implementations with advanced Python libraries - this is beyond my current skill level
Production-grade optimizations and professional software engineering practices
These skills will develop when I enter college and get formal education

My current level: Self-taught, pre-college level. I have practical coding ability, but not at professional or advanced levels yet. This will improve with formal education and experience.

About Privacy and Personal Information

What I shared in this document:

My learning journey and methods
The challenges faced by someone with curiosity but limited financial resources
How I learned and what tools I used (all free and legal)
My hardware limitations and how I worked around them

What I did NOT share:

Any personal information that could compromise security
Any specific details that would enable unethical behavior
Any information about other people without their consent

The Purpose of This Document

This document exists to show:

Financial constraints don't stop curiosity-driven learning - they just change the approach
How someone without resources, money, connections, or mentors can still learn and build
The reality of self-taught learning: it's messy, sometimes wrong, and requires constant self-correction
That AI can be a powerful learning tool when used ethically and responsibly
The challenges of balancing passion with practical requirements (like JEE preparation)

I'm sharing this honestly - including my limitations, mistakes, and uncertainties - to help others who might be in similar situations. This is the reality of learning in isolation with zero resources, driven purely by curiosity that I can't control.

Bottom line: Everything I've done has been legal, ethical, and on my own devices. I'm a learner, not a hacker. I'm curious about how systems work, not about breaking into them.

Project Status: Paused for JEE preparation (December 2025)

Code: github.com/Qeltrix/Qeltrix

License: GPLv3 (code), CC BY-SA 4.0 (concepts)

Community: Fully open for contributions, forks, and improvements

DEV Community