Skip to content
View Amir0234-afk's full-sized avatar
🎓
Studying
🎓
Studying
3 followers · 5 following

Block or report Amir0234-afk

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Amir0234-afk/README.md

header

Typing SVG

LinkedIn

Focus

  • Applied ML (NLP/CV): clean experiments, reproducible training, readable results
  • Security: small utilities + crypto-adjacent scripts
  • Game dev: prototypes + tooling, packaged demos and short postmortems

Pinned projects

  • sentiment-analysis-hf — BERT/Transformers sentiment analysis (IMDB, SST-2): train/eval/batch predict
  • license-plate-segmentation-unet — U-Net 3-class segmentation (TF/Keras)
  • CNN-Experiments-on-NIST-SD19-Handwritten-Characters — modular CNN pipeline (NIST SD19): preprocess → train → evaluate
  • Base-FlappyBirdClone-UnrealEngine-5.1.1 — Unreal Engine Blueprint prototype base

Stack

  • Languages: Python, C++, TypeScript, SQL, Bash
  • ML: TensorFlow/Keras, Hugging Face Transformers
  • Infra / DevOps: Docker, GitHub Actions (CI), PostgreSQL
  • Security/RE (tools I use): WireShark, Ghidra
  • Game Development: Unreal, Godot

Samples (selected evidence)

Selected qualitative and quantitative artifacts from pinned projects. Full analysis in individual repositories.

Sentiment Analysis — Transformers (NLP)

Sentiment training metrics

  • Accuracy: 93.0% | F1: 0.93 | Eval set: 872 samples
  • Common failure modes: sarcasm/irony, stylistic negativity, mixed sentiment
  • Errors often show polarity inversion; low-confidence predictions correlate with ambiguity

License Plate Segmentation — U-Net (CV)

U-Net segmentation output

  • Final metrics (50 epochs):
    Accuracy 0.9982 · Loss 0.0039 · Val Accuracy 0.9974 · Val Loss 0.0087
  • Qualitative mask overlays used for validation
  • License plates intentionally blurred for privacy

Handwritten Character Classification — CNN

CNN confusion matrix

  • Test Accuracy: 0.7835
  • F1: Micro 0.7835 · Macro 0.7832 · Weighted 0.7832
  • Confusion matrix highlights class-level performance variance

footer

Pinned Loading

  1. aml-elliptic aml-elliptic Public

    End-to-end AML detection pipeline on the Elliptic Bitcoin transaction graph, comparing tabular baselines (cuML/sklearn) and GNNs (PyTorch Geometric) with reproducible phases, metrics, and curated r…

    Python

  2. CNN-Experiments-on-NIST-SD19-Handwritten-Characters CNN-Experiments-on-NIST-SD19-Handwritten-Characters Public

    Description Modular pipeline to load, preprocess, train, and evaluate CNNs on NIST SD19 handwritten characters. Supports configurable architectures (filter sizes, kernels, layers, pooling, activati…

    Python

  3. sentiment-analysis-hf sentiment-analysis-hf Public

    BERT-based sentiment analysis for IMDB and SST-2 datasets using Hugging Face Transformers. Includes training, evaluation, and batch prediction scripts under MIT License.

    Python

  4. license-plate-segmentation-unet license-plate-segmentation-unet Public

    U-Net for license plate segmentation (3-class: background, border, plate) in TensorFlow/Keras.

    Python

  5. Base-FlappyBirdClone-UnrealEngine-5.1.1 Base-FlappyBirdClone-UnrealEngine-5.1.1 Public

    This is the base for a Flappy Bird Clone Game made with Unreal Engine Blue Print system

    1