AI Engineer / AI Developer · Amman, Jordan
LLM apps, RAG pipelines, OCR and agent workflows, wired to real backends: FastAPI, Postgres, Docker.
Taken from what is in my public repositories.
Selected work
Chosen from 24 public repositories for what they show about engineering, not for size or stars. Each case study states what is measured, and where nothing is, it says so.
A hiring platform that turns CVs and job posts into structured data, then ranks applicants with an explainable hybrid score.
One FastAPI service that OCRs documents and runs local-LLM summarisation, extraction and question answering, with no cloud AI API.
One vector store, three model backends: ask the same grounded question to a local, a proprietary and a hosted open-source LLM.
A travel assistant that looks facts up in a database through function calling, then answers in text, voice and a generated image.
Smaller tools and prototypes. Source only.
Gradio app that cleans up scanned PDFs and images with OpenCV before OCR, in English, Arabic or both.
LangChain + Chroma chatbot with local MiniLM embeddings, answering only from a source PDF.
Whisper transcription feeding a 4-bit quantised Llama 3.2 3B that writes structured minutes, fully open-source.
Three tone-specific writer agents draft recruiter emails from a CV; a rubric picks one, and sending is opt-in.
University and self-study notebooks, shown with their caveats.
Conv1D, LSTM, GRU and BiGRU compared on 163k labelled tweets.
Validation accuracy: BiGRU 0.9335 (epoch 4), Conv1D 0.924, LSTM 0.920, Flatten 0.903.
SVM and decision-tree baselines on the UCI Parkinson's voice dataset.
SVM: accuracy 0.872, F1 0.923, on a test set of only 39 rows.
Random forest, XGBoost and LSTM on building energy data, explained with SHAP and LIME.
Best random forest RMSE 0.0855 on a target scaled to 0–1.
A Gradio notebook that asks GPT or Claude to translate Python into Java.
Generated Java is not compiled or tested, so no accuracy claim is made.
How I build
A model call is one box in a larger diagram. This is the path from a problem to something people can use.
01
Write down who uses it, what they do today and what 'good' means before touching a model.
Experience
Role, study and dated public work. Employer-specific details stay off the page; the repositories are the record.
Currently working as an AI Engineer.
Built a full-stack hiring platform: FastAPI, async PostgreSQL, Next.js, Claude tool-use extraction and hybrid embedding scoring.
Retrieval pipelines on Chroma, a multi-provider RAG assistant, Arabic/English OCR tools and an offline document-intelligence service.
Tool-calling with multimodal output, open-source speech-to-minutes with a quantised Llama model, prompt-engineering exercises.
Machine learning, deep learning, NLP and robotics coursework. Public notebooks from this period cover sentiment models, classical ML and forecasting.
Foundational certification covering AI workloads and Azure AI services.
Technical map
No percentage bars. Select any skill to see the projects that actually use it.
Making language models dependable inside software.
Foundations from the degree and public notebooks.
The software around the model.
Connecting models to workflows.
On GitHub
Everything featured here is public and linked. The profile has more notebooks and experiments than I show on this page.
Many of the notebook repos are early learning work.
Ask about my work
This assistant reads the same data that powers this page. It can't invent experience: if the answer isn't in the portfolio, it says so.
Pick a question above or type your own.
About
I studied AI and robotics, so I like systems where software meets something that has to actually work: sensors, documents, people, databases.
Most of what I build now sits around language models. A model on its own is a demo. The product is the retrieval, the tool calls, the validation, the API and the interface that make it dependable.
I'd rather ship a small thing that runs and say honestly what it can't do yet than polish a diagram. Every project on this site links to its source, and each case study lists its limits.
Contact
Let's talk. Tell me what you're trying to automate, understand or ship.