Kolkata, India A NOTEBOOK, NOT A FINISHED PRODUCT

Student / programmer / hands-on learner

Taking things
apart.
Making them
my own.

I’m Amitendra, a 13-year-old building across code, circuits, and machine learning. Usually following a question. Occasionally creating three more.

01 / Selected work

Things on
my workbench.

Experiments, prototypes, and ideas that are still finding their shape.

Hardware / In development

Learning
to leave
the ground.

A custom micro quadcopter built around an ESP32-S3. Coreless motors, an MPU-6050, MOSFET drivers, and 3D-printed parts—with plenty to figure out between the code and the workbench.

ESP32-S3
C++
Motor control

Open build notes

The board has 16 MB flash and 8 MB PSRAM. My work includes camera and IMU testing, motor control, and a dedicated flight-control loop. Flight testing and tuning remain part of development.

FIG. 01 / FLIGHT STUDY Illustration, not to scale

AI + security / In development

Ghost

Exploring a multi-agent security research framework with llama.cpp and specialized Qwen models. A project about local inference, inspectable reasoning, and authorized research.

Python
Local models
Security research

Open research notes

I’m exploring how agents can divide research tasks and compare their reasoning. Alongside this, I practice web exploitation and reverse engineering through picoCTF.

The question

Can a group of small models make a useful research partner?

Machine learning / Built & fine-tuned

A small GPT.
A closer look.

A local 20-million-parameter Mini-GPT experiment built and fine-tuned on a laptop. My way of getting closer to architecture, training, and what happens inside a model.

Python
Language models
Fine-tuning

20M Parameters.
Plenty to learn.

Game physics / Experiments

Built for
the next corner.

Racing physics, chassis simulations, and procedural track generators in Roblox Studio. In TORCS, I’m also exploring autonomous driving and how a simulated driver can improve.

Luau
Roblox Studio
TORCS

Ongoing fascination

The gap between a car that moves and a car that feels right.

Computer vision / Trained prototypes

BatchGuard

PCB anomaly-detection experiments trained on Kaggle. Looking for defects while learning how dust, subtle changes, and unfamiliar boards complicate visual inspection.

Computer vision
Kaggle
PCB inspection

The challenge

Knowing which tiny difference actually matters.

Quant research / In development

Jevie

A paper-trading research bot exploring technical indicators, backtesting, and model-driven decisions. A simulated environment for testing ideas and examining their failures.

Python
RSI / SMA / EMA
Paper trading

Research principle

An interesting strategy still needs evidence.

Explore my GitHub

02 / The person

More than
the projects.

I like understanding the systems underneath the things I use.

Sometimes that means writing Python. Sometimes it means wiring a sensor, tuning racing physics in Luau, or working inside Linux on a phone. I move between software and hardware because both give me something different to learn.

A few familiar tools

Languages: Python, C++, Luau
Hardware: ESP32, sensors, motor control
Environments: Linux, WSL, Termux, UserLAnd
Workspaces: Kaggle, Colab, Replit, Codespaces

03 / Off-screen & after hours

Close the laptop.
Keep exploring.

/ 01

A different
kind of rally.

Competitive tennis. Baseline tactics, serve-and-volley, and thinking about the next shot.

/ 02 — Outside

Room to move.

Cycling, movement, and a change of scenery away from the screen.

/ 03 — In-game

One more round.

Roblox Rivals, TrackMania Nations Forever, Rocket League, and Minecraft.

/ 04 — At the board

One move deeper.

Chess is another interest of mine—a different kind of challenge from coding, building, and playing tennis.

04 / Follow along

Still a
work in progress.

Code, experiments, and whatever I decide to understand next.

Find me on GitHub

Taking things apart. Making them my own.

Things on my workbench.

Learning to leave the ground.