I build AI products that check their own work.
2+ years in a marketing department, turning business cases into products that bring the company leads and save the team hours. Most AI demos break the first time the model is confidently wrong. Mine catch it: schemas, validators and QA checks are built in.
I designed, built and run ritorica.com alone, from the Figma file to the Linux server.
Available for new projects · Moscow, UTC+3 · Remote or relocating · English, Arabic, Russian
Work
No demos and no prototypes. Everything here runs, and each project has the numbers behind it. Two here; the rest are one click away, grouped by discipline.

ritorica.com
Ritorica sells business messaging. The site had to explain it to buyers, let marketing edit pages without a developer, and turn visitors into leads. I designed it in Figma, built it on Django CMS with lead capture and a Gemini chatbot, and deployed it to a Linux server with automatic deploys. Alone, and I still run it.
In production since June 2026 · a page per country · English, Arabic and Bengali
Read the case study ↗

AI video and image generator
A brand kit and a topic go in; a finished video, carousel and post copy come out. Claude agents do the thinking: the angle, the script and the final review. The video itself renders locally with Remotion and free neural voices, and a QA step checks it against the script.
€0 in paid APIs · a 26-second video in about 50 seconds on a laptop · English, Arabic or Russian
Read the case study ↗
What I can build for you
Fixed scope, clear timeline, one person from design to deployment. You talk to the person who builds it.
AI chatbot for your website
Trained on your own content. Answers customer questions, captures leads, and hands off to a human when it isn’t sure. Answers are checked against your sources, so it doesn’t make things up.
Company website, designed and built
Designed in Figma, built, deployed and optimised for Google and AI search from day one, the same way I built and still run ritorica.com.
Workflow automation with AI
Reports, emails, data entry or content automated with n8n and AI agents, with checks that catch errors instead of shipping them.
Security check of your web app
OWASP-based review of your website or web app: what an attacker can reach, what to fix first, and a clear written report.
Not sure which fits? Tell me what you’re building and I’ll suggest the simplest option.
What I work with
Five areas, one person. I can build the thing, design it, and explain it to the people who buy it. That is usually three contractors.
-
AI and automation
Agents and workflows with the checks built in.
Claude CodeOpenAI CodexGeminiManusAgent and prompt designJSON SchemaLangChainn8nUiPathChatbots -
Development
Web apps and sites, from the database to the server.
PythonFastAPIDjangoReactTypeScriptHTML/CSSPostgreSQLDockerLinuxBashGit -
Design
Brand systems, interfaces and files ready for print.
FigmaCanvaMiroDesign systemsBrandPrint production -
Marketing
I also know how to sell what I build.
SEOAI search (GEO)Lead generationWhitepapers -
Security
Part-time pentesting, so what I ship starts locked down.
OWASPBurp SuitenmapWireshark
I break things for work, so I build them locked down
I do part-time penetration testing of my company's websites and SMS platform against an OWASP checklist. That work is why the apps below start closed and open only where they have to. Every line here is enforced in code, not promised in a policy.
- AI agents get an allowlist of tools, never open shell access. An agent that can run commands is the attack surface.
- Uploads and file paths are checked, so nobody can read files outside the app. Tests try
/etc/passwdand../and expect a refusal. - The audit app refuses private and internal addresses and re-checks after every redirect, so a stranger cannot make my server fetch my own network.
- Servers run HTTPS and key-only SSH, with hashed passwords and rate-limited sign-in.
- Containers run as a normal user, not root.
- Keys stay in
.env, and every repository is scanned for secrets before it goes public.
Tools: nmap, Wireshark, Burp Suite, OWASP testing, TryHackMe. I also wrote system-audit, a Bash tool that checks a Linux server for the misconfigurations I keep finding.
Experience and education
Avionics engineer by training, then drones, then AI. Based in Moscow, open to remote work and to relocation.
Work experience
-
Sept 2024 to now · remote
AI Marketing Engineer
Ritorica
The website, AI and automation for a B2B messaging platform: SMS, WhatsApp, Viber, Telegram and push. I designed and built ritorica.com, run its SEO and AI-search work, write the technical specs and whitepapers, and handle event campaigns and market-entry research. I also run the company's GSMA membership and its part-time security testing.
-
Feb to Aug 2024
Arabic AI Content Specialist
Yandex
Arabic content for Yasmina, the Arabic version of the Alice assistant, for audiences across the Middle East and North Africa. I reviewed and corrected its answers and reported accuracy problems to the product team.
-
2023
UAV engineer
Moscow Aviation Institute
Drone design and electronics at the university factories, with STM32, Arduino and ArduPilot.
-
Sept 2020 to June 2021
Foreign Students Specialist
Moscow Aviation Institute, International Department
Education
-
2026, in progress
AI and Machine Learning Diploma
Route
Python, machine learning, deep learning, computer vision, NLP and Transformers, LangChain and LangGraph, ending in capstone projects.
-
2025
MSc, Software Engineering
Moscow Aviation Institute
-
2023
BSc, Avionics Engineering
Moscow Aviation Institute
-
Additional
Translation training
RUDN University
Questions people ask first
Short answers for recruiters, hiring managers and clients. Anything else, ask me on Telegram.
Do you take projects or full-time roles?
Both: fixed-scope projects, and longer-term roles in AI product and automation. For a project, pick a service above or tell me what you need.
What does “AI that checks its own work” mean?
The model’s answer passes checks before anyone sees it. A JSON schema checks its shape, validators compare it with your sources, and a QA step can reject it. A wrong answer is caught, not shipped. A chatbot that isn’t sure hands the question to a person.
What can I see that is live?
ritorica.com, which I designed, built and still run alone, from the Figma file to the Linux server. Four technical guides I researched, wrote and designed: 87 pages, live on the same site. Code and tools are on GitHub.
You trained in avionics. Why AI?
Avionics taught me to test everything before it flies. I went from drone electronics (STM32, ArduPilot) to an MSc in Software Engineering, then to Yandex, where I reviewed and corrected the Arabic answers of the Yasmina voice assistant. Since 2024 I build AI products at Ritorica, and I’m taking an AI and Machine Learning diploma.
Can you work in an international team?
That is my daily work. At Ritorica I work with colleagues in Russia, the UAE, Bangladesh and Saudi Arabia, in Arabic (native), English (C1 to C2) and Russian (C1).
Where are you based, and what hours do you work?
Moscow, UTC+3, which overlaps with working hours in Europe, the Gulf and Asia. I work remotely and I’m open to relocation. I reply within 24 hours.
How does a project start?
You send a short message: what you need and by when. I reply within 24 hours with the simplest option, and we agree the scope and timeline before any work starts. You talk to the person who builds it, from design to launch.
Tell me what you are building
Available for new projects
- Replies within 24 hours
- Moscow time, UTC+3: overlaps with Europe, the Gulf and Asia
- English, Arabic or Russian