Software Engineer  ·  M.Sc. Researcher in Artificial Intelligence
Clinical AI  ·  Reliable & Robust LLMs  ·  AI-Enabled Software Systems
_{Mashhad, Iran}

I am a software engineer and M.Sc. student in Artificial Intelligence at Islamic Azad University, with 4+ years of experience in backend engineering, distributed systems, DevOps, and applied machine learning. I lead backend and AI systems development for a U.S. medical toxicology practice, where I own ToxLumen, an AI-enabled clinical toxicology platform.

My research examines the reliability and robustness of AI systems in clinical decision-making, with a particular focus on tool-using large language models.

M.Sc. Thesis (in progress) — Decision-Boundary Robustness in Tool-Using Clinical LLMs: A Counterfactual Benchmark in Clinical Toxicology
Islamic Azad University · Supervisor: Dr. Yahya Forghani · Advisor: Dr. Omid Mehrpour (Medical Toxicologist)

• Question: Do tool-using clinical LLM agents recognize when a small but clinically decisive variable crosses a meaningful decision boundary, and change their decisions accordingly?
• Method: A structured counterfactual benchmark of minimal case pairs in clinical toxicology — each pair holds the clinical context fixed and alters a single decision-relevant variable.
• Evaluation: Boundary recognition, unsafe use of invalid tool outputs, triage behavior, and next-step routing — together with runtime guards and structured decision reporting, to determine whether agents fail safely, fail silently, or correctly redirect the clinical workflow.

• Clinical AI & Decision Support — safety, validity, and clinically meaningful behavior of AI systems that support or evaluate clinical decisions.
• Reliable & Robust Large Language Models — decision-boundary behavior, counterfactual sensitivity, and the safe use of tool-augmented models.
• AI-Enabled Software Systems — engineering production software that integrates ML and generative AI into real-world workflows.
• Distributed Systems, MLOps & Cloud Infrastructure — reproducibility, asynchronous processing, containerization, and production ML systems.

• Medical Toxicology — Lead backend and AI systems; designed the multi-service architecture behind ToxLumen (intake, transcription, LLM-assisted analysis, clinical notes), consolidated four applications into a single product, and run containerized production deployments while leading the engineering team.
• Sportera — Co-founded a fitness platform and own its technology: Django/DRF backend, a training domain built on immutable session snapshots, gym-owner and admin frontends, and an ongoing migration to NestJS and Prisma.
• Niken 3D — Own the API backend and production infrastructure for a 3D-printing e-commerce platform, including CI-gated deployments with test-coverage enforcement, observability, datastore hardening after a ransomware incident, and off-site backups.

• ToxLumen — AI-enabled clinical toxicology platform (ongoing). Multimodal case intake, transcription, structured extraction, and LLM-assisted analysis; three applications on a shared Django backend with organization-scoped billing. Guiding principle: AI assists; the clinician decides.
• FaceCheckIn — IoT attendance and affect-logging system (academic). Raspberry Pi, identity recognition, image-based inference, secure APIs, PostgreSQL, and a mobile client. Top-rated project (full marks) in Fundamentals of IoT.
• University Course Scheduling System — Academic. Automated class scheduling and an administrative dashboard for Quchan University of Technology, with manual override and Excel import.

• M.Sc. in Artificial Intelligence — Islamic Azad University · Feb 2025 – Present
• B.Sc. in Computer Engineering — Quchan University of Technology · Sep 2020 – Jun 2025 · GPA 16.37 / 20

• Machine Learning — Teaching Assistant, Islamic Azad University (Dr. Yahya Forghani), Autumn 2025
• Advanced Programming (C++) — Teaching Assistant, Quchan University of Technology (Dr. Zahra Eskandari), Autumn 2023
• Basic Programming (C) — Teaching Assistant, Quchan University of Technology (Dr. Malihe Javidi), Spring 2022

Area	Tools
Languages
AI & Machine Learning	OpenAI API · Gemini API
Backend	Django REST Framework
Data & Messaging	Celery · Huey
DevOps & Cloud
Frontend & Mobile	React Native

_{Languages: Persian (native) · English (advanced)  |  Full profile, research and CV at sjdfani.com}