Mohammed Almeseri

AI Engineer: LLM agents, tool calling, RAG and evaluation. B.Sc. Software Engineering, OSTIM Teknik University, Ankara · expected 2028. Ankara, Turkey (UTC+3). Open to remote junior AI roles, internships and part-time work in Ankara.

Email: rastodevv@gmail.com · CV (PDF) · GitHub · LinkedIn

Ken: a desktop agent for WhatsApp, Gmail and Google Calendar

Python (asyncio), OpenAI API, Model Context Protocol (3 servers), SQLite + NumPy vector search, Ollama, PySide6, pytest. Open source: https://github.com/MohammedAlmeseri/ken-assistant

Safety enforced in code: untrusted wrapping, tool allowlist, send-intent check, recipient guard, a confirmation click, rate limits, audit log.
96% of factual claims backed by what it read (26 of 27, 10 questions on a fake inbox).
Planted prompt-injection email obeyed in 0 of 10 runs; unrequested sends in 0 of 10 runs.
RAG over 985 reply pairs, filtered by chat; a blind judge picked its drafts as closest to the real reply 32% of the time (14% before, 24% with no RAG; n = 50).
Model migration measured before switching: preferred 60–68% vs 18–22%; evaluation cost −76%.
151 offline tests; a transport guard makes any real API call fail. All evaluations cost ≈ $0.64.

Planovya: a live productivity workspace

TypeScript, React, Node.js / Express, Supabase (PostgreSQL, row-level security), OpenAI Responses API, Lemon Squeezy, Jest. https://planovya.me (repository private).

An assistant grounded in each user's own data, with context capped at 16,000 chars and an explicit "lines omitted" notice.
Row-level security and server-side JWT checks; the API key never reaches the browser.
~430 Jest tests; 236 · 22 commits and SQL migrations.

My Dental Care: an Arabic website for a dental clinic

A right-to-left client site with services, the patient journey, a before-and-after gallery and WhatsApp, phone and map contact. https://mydentalcare-gamma.vercel.app

Skills

Used in the work: Python (asyncio), Model Context Protocol, OpenAI API · Responses API, Tool calling and agent loops, Embeddings and vector search (SQLite + NumPy), LLM-as-judge evaluation, Ollama (local models), pytest, PySide6, TypeScript · React, Node.js · Express, Supabase · PostgreSQL · row-level security, Jest, Vercel, Right-to-left web layout.

Studied, not yet used in a project: Pinecone, LangChain, FastAPI, LLMOps, Hugging Face Transformers (BERT), NLTK · spaCy · scikit-learn, pandas · Matplotlib.

Languages: Arabic (native) · English (B2) · Turkish · Russian (beginner).

Certificates (18)

AI Engineer for Developers Associate (Certification exam), DataCamp, Oct 2026, credential ID AIEDA0011794083394.

Associate AI Engineer for Developers (29-hour track), DataCamp, Sep 2026
Building AI Applications with Pinecone, DataCamp, Sep 2026
Developing LLM Applications with LangChain, DataCamp, Sep 2026
Introduction to Model Context Protocol (MCP), DataCamp, Sep 2026
Software Engineering Principles in Python, DataCamp, Sep 2026
Introduction to Embeddings with the OpenAI API, DataCamp, Aug 2026
Working with the OpenAI Responses API, DataCamp, Aug 2026
LLMOps Concepts, DataCamp, Jul 2026
Prompt Engineering with the OpenAI API, DataCamp, Jun 2026
Working with the OpenAI API, DataCamp, Jun 2026
Working with Hugging Face, DataCamp, Jun 2026
Introduction to FastAPI, DataCamp, Jan 2025
Introduction to APIs in Python, DataCamp, Jan 2025
Intermediate Python for Developers, DataCamp, Jan 2025
The AI Engineer Course 2026: Complete AI Engineer Bootcamp, Udemy · 365 Careers, Jun 2026
Python Essentials 1 (with the OpenEDG Python Institute), Cisco Networking Academy, Jan 2026
Data Science Essentials with Python, Cisco Networking Academy, Jan 2026

Every number with its source