Amal Baby Mathews
SYSTEM ONLINE

Amal Baby
Mathews

ENGINEER // AI & ML

01 // IDENTITY

THE PHILOSOPHY

Hi! I'm Amal. I speak tech, but I'm just a human figuring things out. Beyond code, I love hitting the gym, sci-fi movies, and deciphering the constantly changing code of NLP.

[ + Click to Expand ]

I operate at the intersection of robust engineering and experimental AI. My work isn't just about training models; it's about architecting systems where traditional software reliability meets the stochastic nature of Large Language Models.

> User.status: "Optimizing..."
> User.drive: "Maximum"
02 // ARSENAL

CORE FOCUS

Specializing in bridge-building between classic software paradigms and modern AI behaviors.

ENGINEERING

  • Software Engineering
  • Python & Django
  • Web Architecture

AI / ML

  • LLMs & RAG
  • Computer Vision (VLM)
  • NeuroEvolution (NEAT)

Technical Depth:

  • Full Stack AI: Integrating Vector Databases (Faiss/pgvector) with Django backends.
  • Vision Pipelines: From OpenCV pre-processing to VLM inference.
  • Cloud & Deploy: Containerization and scalable inference endpoints.
[ + Click to Expand ]

SELECTED WORKS

Django Web App

Django Web App

RAG Django Faiss DB

Next-gen web app with personalized workspaces and AI chat.

[ + Click to Expand ]

Repository: djnappnew | Architecture: Modular Django Monolith

  • Aweapp/: Core logic handling LLM interfacing and prompt engineering.
  • fileupload/: Decoupled app for ingesting user documents (PDF/TXT) before indexing.
  • register/: Isolated authentication system separating user management from AI logic.
Pong Automation

Pong Game Automation

RL NEAT Python

Self-playing Pong game using NEAT to evolve optimal paddle behavior.

[ + Click to Expand ]

Repository: Pong Game ML Repository | Stack: Python, NEAT-python

  • Historical Markings: Tracking gene origin to prevent redundant structures.
  • Speciation: Grouping similar networks to encourage diversity.
  • Incremental Growth: Starting simple and adding complexity over time.
VLM Test

VLM Moondream Test

Computer Vision Streamlit OpenCV

Advanced analysis using Moondream AI 0.5 with object overlays.

[ + Click to Expand ]

Repository: VLM_testing | Stack: Streamlit, PyTorch, OpenCV

  • Benchmarking: Parallel testing of Moondream, SmolVLM, and Janus models.
  • Visual Feedback: Programmatic drawing of bounding boxes on video frames using OpenCV.
  • UI: Streamlit-based interface for rapid prototyping.
Stock Prediction 1 Stock Prediction 2

Stock Trend Prediction

NEAT Finance Time Series

Exploratory project using NEAT to predict stock trends (TSLA) via evolution.

[ + Click to Expand ]

Repository: NEAT Stock Prediction | Stack: Python, NEAT-python, Pandas

  • Input: Previous day's OHLC (Open, High, Low, Close) data.
  • Method: Evolving neural network architectures automatically to find optimal configurations.
  • Skills Applied: NEAT principles, data preprocessing/feature engineering for finance.