Applied AI Intern · Dentist (BDS) · Healthcare MBA Candidate
Egypt · Open to remote opportunities
Reliable agentic AI for high-stakes systems. I build reviewable workflows around tool permissions, source provenance, failure analysis and human judgment, informed by my work in dentistry and clinic operations.
I am a practicing dentist and dental clinic owner based in Egypt, with a BDS from Modern Sciences and Arts University and an MBA in Hospital and Healthcare Facilities Administration in progress. Alongside clinical and operational work, I am building hands-on fluency in Claude, prompt engineering, generative AI, data analysis, and AI-assisted workflows.
Working code, reproducible checks, and visible limitations. Each project distinguishes synthetic contract checks from real-world performance.
FastAPI and React dashboard, strict prediction contracts, trace scoring, failure taxonomy and regression reports. Scripted controls score 50/100 and 100/100; this measures evaluator behavior, not LLM capability.
Fixed skills, explicit workflow states, role allowlists, bounded steps, repeated-step detection and trace export. A single-use local approval seam binds to the exact proposal; production reviewer authentication remains pending.
Agent summaries carry provenance metadata. Cross-scope memory-ID overwrites are rejected, and tests cover session visibility, TTL expiry and durable reopen. Comparable database benchmarks remain pending.
An image-quality research baseline with documented exact-image grouped folds. New split validation rejects missing patient IDs and leakage. Expert labels, patient-disjoint analysis and external clinical validation remain pending.
Genopedia binds lexical retrieval to source URLs, versions, license labels and SHA256 hashes. Neurapedia adds Dice/IoU, calibration and error-count metrics. Current new fixtures test contracts; no diagnostic claim.
Existing supervisor exercised by healthy control and seven faults: sensor dropout, latency, pose dimensions, workspace limits, unauthorized controller, emergency stop and command replay. No hardware or tissue execution.
Arab Academy for Science, Technology and Maritime Transport
Modern Sciences and Arts University
Anthropic Academy · Jul 2026
Project Management Institute · Cisco Networking Academy
The current evaluator uses ten task templates with ten fixture variants each. The v1 and v2 agents are deterministic scripted controls with explicit tool requests. A perfect control score verifies the fixture path; it does not establish autonomous AI capability.
| Observed metric | v1 | v2 | Eligible fixtures |
|---|---|---|---|
| Task success | 50/100 | 100/100 | 100 |
| Citation accuracy | 0% | 100% | 20 |
| Transient recovery | 0/10 | 10/10 | 10 |
| Unauthorized attempts | 10/100 | 0/100 | 100 |
| Success regressions | — | 0 | 100 |
Inspect the full report → · Download JSON · Reproduce from source
79/256 exact answers on its synthetic held-out test; 177/192 non-release cases were incorrectly released. The rejected adapter and its limits are documented. Real agent, clinical and hardware validation remain open research gates.
Open to remote applied AI roles, internships and collaborations in agent evaluation and healthcare technology.