AI Engineering

Automation that saves hours.

I build AI systems that take repetitive work off people's desks: grading, scheduling, document handling and reporting.

Services

LLM assistants

Language models tuned or prompted on your own documents to answer, classify or grade text.

Document extraction

PDFs, scans and spreadsheets turned into clean, structured records.

Scheduling and optimisation

Timetables, rosters and assignments solved with constraint programming.

Workflow automation

Reports, emails and data entry handled by scripts and AI agents.

Projects

01

Automated essay grading

Problem: Marking thousands of written answers by hand is slow and inconsistent.
Built: A Gemma 2 9B model fine-tuned with LoRA on 2,550 accounting and auditing essays, scoring four rubric criteria.
Result: Benchmarked against GPT-4o, Claude and three human examiners. Code, model and dataset are public.

LLM
LoRA
Python

GitHub ›Hugging Face ›Zenodo ›

02

Exam scheduling solver

Problem: Assigning supervisors and TAs to every exam slot and room by hand is slow and error-prone.
Built: A constraint-programming model that places each person under availability, load and room rules.
Result: Supervisor and TA assignments generated automatically for the college's exam periods.

OR-Tools
CP-SAT
Python

03

Sections timetable system

Problem: Building TA section schedules from registrar reports meant retyping data and catching clashes by eye.
Built: A pipeline that reads the registrar PDF, builds the timetable, flags clashes and exports a PDF per TA.
Result: Used for the Information Systems department's Fall 2026/27 sections: 57 lecture groups, 9 TAs.

Python
PDF parsing
Automation

Contact

Tell me which task takes your team the most time. I will reply with whether it can be automated and how.

Email: kamal.ezzo@must.edu.eg
Education: MSc Software Engineering, Nile University, 2026
Focus: LLM fine-tuning, automation, optimisation