AI engineer
I build fast, secure software and the models that run inside it — from Transformers on streaming sensor data to hardened web apps shipped end to end.
What I reach for to take an idea from notebook to production. Honest tiers: the bottom row is what I'm learning right now.
Next.js
Production-ready
React
Production-ready
TypeScript
Production-ready
Tailwind CSS
Daily driver
Three.js
Production-ready
Flutter
Production-ready
Python
Daily driver
FastAPI
Daily driver
Flask
Production-ready
Django
Production-ready
PHP / Laravel
Production-ready
Node / Express
Production-ready
Go
Levelling up
MySQL
Production-ready
MongoDB
Production-ready
Linux
Daily driver
Git
Daily driver
Docker
Levelling up
Kubernetes
Levelling up
Kali toolchain
Levelling up
Web, security and AI projects. Each card shows where it stands: building, shipped, or still a concept. Open one for the details.
Attack-surface recon that tells you what changed.
A dashboard that combines subfinder, amass and httpx results, stores them per scope, and re-scans to flag new subdomains automatically. Filter to what changed and export it to your next tool.
Attack-surface recon you can actually read.
A self-hosted recon console that runs Nmap and service fingerprinting, maps exposed hosts into a live graph, and matches versions against public CVE feeds — so a scan becomes a ranked to-do list instead of a wall of text.
Links coming soon
Secrets that burn after reading.
Zero-knowledge secret sharing. Payloads are encrypted in the browser with AES-GCM, the key never leaves the URL fragment, and the ciphertext is destroyed after one read or a timer. The server never sees plaintext.
Links coming soon
A firewall for LLM prompts.
A drop-in proxy in front of any LLM API. Every request is scored for prompt injection, jailbreak patterns and data exfiltration before it reaches the model, and a dashboard shows what was blocked and why.
Repository opens when work starts
My thesis model, shipped as a mobile SDK.
A Flutter plugin that runs a parameter-efficient Transformer on the phone's IMU stream and emits activity events in real time. No server, and no sensor data leaves the device.
Links coming soon
Watch the internet knock on your door.
A fleet of containerised low-interaction honeypots (SSH, HTTP) that stream every login attempt and payload to a live 3D globe, turning background noise into threat intelligence you can see.
Repository opens when work starts
Distributed training from one dashboard.
A control plane for PyTorch DDP jobs on Kubernetes: launch multi-GPU runs from a web UI, stream logs and metrics live, and resume from checkpoints automatically when a node disappears.
Links coming soon
Research labs, classrooms and the projects in between.
Now
Independent
Shipping PhantomScope, VaultLine and Forgeline in public while going deep on Docker, Kubernetes and the Kali toolchain.
2023 – 2024
Biomedical and Neuroscience Research Laboratory
Maintained and calibrated CT and MRI equipment, and built EEG collection and analysis pipelines for Alzheimer's diagnosis.
2020 – 2023
NAIRG Lab, Shahid Beheshti University
Designed a lightweight Transformer for on-device activity recognition and built Wi-Fi CSI preprocessing pipelines. Four papers came out of this work.
2020 – now
Private tuition
Six years of explaining hard things clearly, one student at a time.
2018 – 2022
Sahand University of Technology, Shahid Beheshti University
Digital signal processing, telecommunication networks, computer networks and convex optimisation.
Roles, collaborations, or a system that needs to be fast and hard to break. I reply within a day.