Hey, I'm Oussema Ben Ameur.

2nd-year Applied Computer Science Engineering student at The National Engineering School of Sousse. I enjoy building things across the full software spectrum, with a growing focus on machine learning & data science.

~ / gnar

gnar@oba:~$ now

→

CurrentlySoftware Engineer · Part-time · Core Techs Solutions, Inc.Working across data science, machine learning, and software engineering — from building and shipping features to training models and making sense of data.

Scroll

01 — About

What I'm into

A short list, in my own words. Not a mission statement.

Machine learning — model building, experimentation, and understanding the math underneath.
Data science — wrangling messy data into something meaningful.
Software in general — I'll try almost anything once.

Currently working on

Java–C++ interoperability for OCR systems

Currently exploring

Scientific ML and high-performance computing

02 — Work

Where I'm shipping

Part-time, in the thick of it. Document intelligence and computer vision at Core Techs Solutions.

Part-time
Part-Time Software Engineer
Core Techs Solutions
September 2025 — Present · Sousse, Tunisia
- —Built an end-to-end document extraction pipeline combining PaddleOCR, LayoutLM, and Qwen-VL with a RAG system for intelligent document querying.
- —Engineered a C++ inference layer with Java FFM bindings to run PaddleOCR and LayoutLM natively in the JVM, eliminating Python dependency overhead.
- —Built and refined deep learning models in PyTorch for production computer vision applications.
- —Shipped features across mobile (Flutter) and backend (Spring Boot) services for company products.
Internship
Software Engineering Intern
Core Techs Solutions
June 2025 — July 2025 · Sousse, Tunisia
- —Designed and trained a computer vision model to assess physical document condition — detecting folds, tears, and degradation for automated quality control.
- —Built data preprocessing and feature extraction pipelines in PyTorch, improving model accuracy over baseline.

03 — Projects

Things I've built or broken

A few that taught me something worth keeping.

OCR System for Financial Records

JANUARY 2026

Problem. Manual processing of financial records and invoices is slow, error-prone, and inefficient, making it difficult to accurately extract and manage key financial data.

What I took away. Ongoing

Book Status Assessement

NOVEMBER 2025

Problem. Assessing the physical condition of used books manually is subjective, time-consuming, and inconsistent, especially at scale for resale platforms or libraries.

What I took away. Applied computer vision to real-world data, handled noisy and imbalanced datasets, optimized model performance through hyperparameter tuning and deployment formats (ONNX, TensorRT), and gained experience in data preprocessing and evaluation.

Pytorch
Optuna
Python
NumPy
Pandas
Matplotlib
FastAPI

AI-Powered Writing Assistant

OCTOBER 2025

Problem. Drafting emails and meeting summaries from rough bullet points, turning them into something readable without losing the writer's tone.

What I took away. Fine-tuning an LLM on business communication data, and the gap between raw model output and something a real person would send.

Spring Boot
React
Flutter
Fine-tuned LLM

FallahTech RAG

APRIL 2026

Problem. An investment committee evaluating a Tunisian AgriTech startup's Series A needs to query a dense, heterogeneous document corpus — financials, legal statutes, market studies — accurately and without hallucination, citing exact figures and refusing when information is absent.

What I took away. Retrieval quality drives answer accuracy far more than model size: hybrid semantic + BM25 + RRF fusion, semantic prefixes, and a hard confidence threshold are what keep the system grounded. Between the two implementations, the Python/LangChain pipeline is the more serious tool — persistent storage, query rewriting, and hybrid retrieval — while n8n trades depth for ease of deployment.

LangChain
ChromaDB
Groq
BM25
FastEmbed
n8n
Ollama
PyMuPDF
openpyx

Play Store Market Analysis

DECEMBER 2024

Problem. Predicting install counts across a 3M-row dataset of Play Store apps, and prototyping a market-research tool for indie developers.

What I took away. Handling scale with pandas and Keras without the notebook falling over, and how much cleaning matters before the model sees anything.