I work at the intersection of software engineering, applied AI and product development—from cross-platform applications and enterprise systems to computer vision, data analysis and nano-bubble water-treatment technology.
I am a software engineer, inventor and founder focused on turning complex problems into dependable, maintainable products.
I am Sepehr Fathi. Since 2021 I have worked professionally across software engineering: cross-platform Flutter applications, PWAs, administrative platforms and web systems. I am the senior developer of the Bardaskan citizen super app for Android and the web, and I founded Vira Web Aria, where I now guide product and technical direction as chairman.
Alongside entrepreneurial work with Shenasnameh Khodro, Masraf and AvaNobat, I am CEO of Nano Hobab Roshana and inventor of a nano-bubble water-treatment system. I also work as a senior software engineer with Cuatro and NanoMAR in Norway, developing applied-AI, computer-vision, geospatial-risk and bioinformatics systems.
Founding, chairing, leading and building — across product, engineering and R&D.
I founded Vira Web Aria — the company behind the Onwebs digital agency (web, apps, design, SEO & marketing) and its product R&D — and chair its board.
Mashhad, IranCEO of the company commercialising the nano-bubble water-treatment device I invented.
IranCo-founder of several product startups — a vehicle-identity platform, a consumption-tracking app and an appointment/booking app.
IranSenior developer building AI-driven and geospatial platforms — including a Norwegian property-risk system — for two technology companies.
NorwayTechnical leadership and development of Bardaskan's citizen-services super app for Android and the web, bringing municipal services, public information and citizen communication into one cross-platform product.
Bardaskan, IranAn engineering foundation spanning mechanics, software and technology.
Mechanical-engineering studies centred on analysis, design and systems thinking—the same foundation I apply to product architecture and technical problem-solving.
Mashhad, IranComputer studies complementing my professional path in software development, systems design and intelligent technologies.
IranSecondary education in Iran's gifted-student programme, where analytical thinking, problem-solving and an interest in engineering took shape.
Mashhad, IranA few products that show the range — each one real and shipped.
The online appointment platform for doctors and clinics in Mashhad that I founded — from patient search to the secretary and admin panels, with a dedicated SEO-optimized page for every physician.
A citizen super app for Bardaskan Municipality: municipal services, announcements and the channel between residents and the city, in a single app.
The internal calendar and automation platform at Onwebs: a PWA calendar, task board, notifications, call centre and native widgets.
A digital file for every car: technical work and repair history kept in one online record that transfers cleanly to the next buyer at sale time.
A platform that reads building and roof condition from satellite imagery and turns it into risk and financial insight for insurers and banks. I build the product front end.
An on-premise smart car-finder for parking lots that reads Iranian license plates from existing cameras and shows a visitor exactly where their car is — with a photo — in seconds, running entirely on one mini-PC inside the building (a real ALPR kiosk with a LAN admin panel).
A YOLOv8-based object-detection model that spots flying objects — birds, airplanes and other aerial objects — in images, recorded video and a live webcam feed, wrapped in a local web app that runs entirely on your own machine.
A nano-bubble water-treatment technology I invented — chlorine-free water disinfection using nanobubbles — commercialised through Nano Hobab Roshana (Iran) and NanoMAR AS (Norway). Physical deep-tech taken from idea to product.
A full-stack system that identifies and ranks candidate target antigens for CAR-T cell-therapy design from real NCBI GEO genomic data — a rigorous, hypothesis-generating research tool with an AI interpretation layer.
A PHP platform to compare, price and resell multi-model AI (LLM) messaging — with a user wallet and payment gateway — turning access to several models into a sellable product.
Websites and platforms delivered under Onwebs (Vira Web Aria), the studio I founded.
I treat prompt engineering as the design of the interaction layer between models, data and software—not as writing a clever instruction. That layer must make model behaviour measurable, controllable and maintainable before AI can become a dependable part of a real product.
I define the model's role, objective, constraints, examples and source priorities explicitly, then supply only the context relevant to the task.
Structured Outputs and JSON Schema turn model responses into reliable software contracts, backed by validation, retries and explicit fallback paths.
Tool and function calls use least privilege, auditable traces and confirmation for sensitive actions; the model coordinates while deterministic code enforces rules.
RAG, targeted retrieval, citations and freshness checks connect answers to verifiable evidence and make missing information explicit.
Prompts, models and settings are versioned like code. Golden datasets, edge cases and regression tests measure quality, latency and cost before release.
I design defences against prompt injection, data leakage and out-of-scope use, with human review and safe stopping points for consequential decisions.
The goal is not one impressive answer; it is stable, measurable and responsible behaviour at production scale.
What I actually work with, day to day.
A wider sample of the sites, systems and panels I've delivered.
Some of the code is public on GitHub, and I'm happy to walk through any of it on request.
Sepehr Fathi is a software engineer, inventor and founder based in Mashhad, Iran. He founded and chairs Vira Web Aria, leads Nano Hobab Roshana, and works as a senior software engineer with Cuatro and NanoMAR in Norway.
Sepehr's personal portfolio covers web and mobile engineering, Flutter, Python and FastAPI, applied AI, computer vision, data analysis, geospatial systems, bioinformatics and nano-bubble technology.
His work includes instruction and context architecture, Structured Outputs, tool calling, RAG and grounded retrieval, evaluation and regression testing, and defences against prompt injection.
For software, applied-AI or research collaboration, contact Sepehr at ftsepi@gmail.com or use the LinkedIn, GitHub and Telegram links on this page.
A custom site or system, an interface that needs rethinking, or a project that has to be found in search — I'd like to hear about it.
