Available — taking on new projects
Python developer with 17 years of experience. I build AI assistants on your data, automate requests and emails, launch bots in Telegram and Max, and connect them to your CRM and website.
Not a “ChatGPT wrapper”, but a system with logs, metrics and an SLA.
Tasks
I don’t start by picking a model but with the process: where time, money and customers are lost. These tasks most often pay off first.
A good AI project pays off not with a pretty demo but with the hours employees stop spending on routine work.
Result
~55%
of requests closed without an operator
How it works
Result
−80%
manual CRM data entry
How it works
Result
−80%
errors in company details
How it works
Result
10–30 s
per answer instead of 20–40 minutes of searching
How it works
Result
80%
of routine reports without an analyst
How it works
Result
~45%
of bookings without an administrator
How it works
The figures are results of real projects under NDA. Outcomes depend on your data and processes, so I give an estimate after analyzing the task.
Don’t see your task? Describe it — I’ll suggest a solution and a timeline.
Describe your taskServices
Prices are the starting point for a small project. I give the exact cost and timeline after analyzing the task and fix them in the contract.
I’m responsible not for what “the chat wrote” but for how the system works for you every day: with tests, logs and a clear handover to your team.
Main focus
AI
The agent answers customers and staff using your data, works with your CRM and APIs, and hands difficult cases to a human.
from $350Pilot in 5–10 days
You get: An agent in your channel: website, Telegram, Max or internal chat
Messengers
A bot or Mini App for bookings, orders, payments and broadcasts — connected to your CRM, not “buttons in a vacuum”.
from $150Launch in 3–7 days
You get: A working bot, documentation and code handover
Integrations
I connect your CRM, website, accounting, payments and messengers so that nobody copies data by hand.
from $500Launch in 5–14 days
You get: Data exchange that doesn’t lose data and reports failures
Prompt audit and rewrite, an eval set, structured output
from $150
per estimate →
Workflows with webhooks, schedules, CRM and Google Sheets
from $150
per estimate →
Scripts, scrapers, FastAPI and Django services
from $350
per estimate →
Next.js, React, Django: customer accounts, SaaS, corporate websites
from $350
per estimate →
Architecture, answer quality, token costs
$45/hr
min. 10 hours
Themes, plugins, WooCommerce, speed-ups and migrations
from $350
per estimate →
Per estimate
Fixed price and timeline; stages and acceptance are set out in the contract.
$45/hrmin. 10 hours
Improvements, consulting, prompt and architecture audits.
$4,000/mo
Full-time work in your team on a product or an area.
Demos
Bots, AI agents, apps and websites for fictional businesses. These aren’t mockups: everything runs on the same infrastructure as client projects — open them and try them before writing to me.
Anyone can say “I can do it”. That’s why I show working systems: open them and check for yourself, and I’ll answer for what’s inside.
56 live demos5 bots in Telegram and Max · 38 web apps · 13 websites
Web app · agent diagram
A support agent is assembled from blocks: intents, a knowledge base, API calls, checks and operator handover. Send a customer message and watch its path light up on the diagram step by step.
▶ Open live demo
Web app · speech recognition · neural network in the browser
A customer dictates a request by voice, and the manager gets a ready card: name, company, phone, budget, timeline, task and a score — hot lead or not. Speech is recognized by Whisper right in the browser.
▶ Open live demo
Telegram Mini App · payment · statuses
A catalog as an app inside Telegram, a cart, test payment and order status notifications — from “packing” to “delivered”.
▶ Open live demo
Approach
I work in stages, each with a clear outcome. You see the progress and can stop after any stage — everything done stays with you.
The most expensive mistakes are made before the first line of code. That’s why I don’t spare time on analyzing the task and the architecture.
I study the process and the data and estimate whether AI will pay off. If the task can be solved without a neural network, I’ll say so.
Deliverable: Solution diagram and estimate
I choose models, data sources and integrations, decide where data is stored and what stays with people.
Deliverable: Technical design and stage plan
I build a working scenario on your data — not a presentation, but a system you can try.
Deliverable: Working scenario on a test environment
I run a set of reference questions and documents, measure accuracy and find errors before your customers do.
Deliverable: Answer quality report
I deploy to your server, connect messengers and the CRM, and show your team how to work with the system.
Deliverable: A working system and documentation
I watch metrics and logs, expand the knowledge base, and update models and prompts as the load grows.
Deliverable: Metric reports and improvements
Timelines are typical for a mid-sized project. I give an exact plan and estimate after analyzing the task.
Time goes to the task, not to an imitation of it
Calls, screen monitoring, always-online mode, night work and free test assignments distract from the work, burn time for nothing and hurt productivity and the result.
A call takes an hour out of the working day: preparation, the conversation and getting back into the task. And spoken agreements get lost and later surface as disputes.
Instead
Written communication in Telegram: questions and decisions are on record, and you can always go back to them.
Screenshot trackers control the process, not the result, and get in the way of focus: you work for the camera, not for the task.
Instead
A transparent result: the repository, a test environment and a short status for every stage.
Being ready to reply within a minute chops the day into pieces, and a complex task never gets hours without distractions.
Instead
I reply on working days and handle messages in batches — that way deep work goes faster.
At night the number of mistakes grows, and fixing them later eats up more time than was gained.
Instead
I work during the day, UTC+5 (Moscow time +2), with a clear head.
A test assignment shows how a practice task gets solved, not yours, and takes days away from current projects.
Instead
A whole section of live demos and a portfolio of projects: open them and check for yourself.
See the demosReviews
Reviews from FL.ru and Upwork, unedited, as clients left them after the deal was closed. FL.ru reviews are translated from Russian; the original is one click away.
The best rating of my work is when a client comes back with the next task.
Latest review
An excellent specialist. He built the platform’s complex functionality quickly and without unnecessary questions. I was very pleased with the attention to detail in UI/UX, the smooth performance of the interface and the instant feedback when solving any technical problem. I’m 100% satisfied with the work, and we will continue working together.
About
I started programming at 21 and have spent 17 years in commercial development. I’ve worked at IT companies in Russia and abroad and twice ran my own business. My clients come from 25 countries.
I started with websites: WordPress themes and plugins, WooCommerce stores, high-traffic editorial projects. Today my main focus is AI systems: agents and assistants on company data, Telegram and Max bots, integrations with AmoCRM, Bitrix24, Yougile and websites. Alongside that, I’m studying at Moscow State University’s Faculty of Computational Mathematics and Cybernetics, majoring in “Developer of Artificial Intelligence Systems”.
AI does not replace architecture. I set the task myself, cut what isn’t needed, read every change in the code and take responsibility for how the system works for you after launch — not for whatever “the chat wrote”.
The task first, then the technology
I dig into the process and the goal and cut what isn’t needed. We build what pays off, not everything neural networks can do.
Architecture and responsibility are on me
AI speeds up routine work, but the problem statement, code review and how the system runs after launch are my job.
Always in touch, no surprises
I run long projects in milestones with a short status update for each stage. You always know what’s done and what’s next.
I hand over a system, not code “to have a look at”
A clear admin panel, documentation and access credentials — your team uses the result without me.
FAQ
About money, timelines, data and what happens after launch.
If a task can be solved without a neural network, I’ll say so. Unnecessary AI means unnecessary costs and unnecessary risks.
Starting prices are in the “Services” section: bots from $150, AI agents from $350, API integrations from $500. The total depends on the number of data sources and integrations, the required answer quality and where the system will run. I give an exact estimate after analyzing the task and fix it in the contract. Improvements and consulting are billed hourly at $45/hr, minimum 10 hours.
A mid-sized project goes from task analysis to launch in 2–5 weeks: analysis 1–2 days, architecture 2–3 days, MVP 1–3 weeks, quality check and launch about another week. I give an exact stage plan after analyzing the task.
The system runs on your server: the database, documents, logs and access keys stay with you. I work with data only within the project and under an NDA. If data cannot be sent to external services, I use models deployed on your own server.
OpenAI and Claude when answer quality and speed matter; open models on your server when data must not leave it. I choose the model for the task: I compare options on your examples and show accuracy and cost in numbers. Stable API access is built into the architecture from the start.
Mistakes are planned for in the architecture. Answers are based on your sources, guardrails forbid making up prices, deadlines and company details, and a difficult case goes to a human together with the conversation history. Before launch I run a set of reference questions and measure accuracy; after launch I watch the metrics and review failed answers.
I can maintain the system: watch metrics and logs, expand the knowledge base, update models and prompts. The format is hourly or a dedicated developer. The code, documentation and access stay with you, so your own team can maintain the system too.
Yes, as a registered sole proprietor: a contract and closing documents, and I sign an NDA before work starts. Stages, deadlines and acceptance are set out in the contract. If you prefer, we can work through Safe Deal on FL.ru.
Yes, and most often that’s the right move. We take one process — for example, answers to typical questions or sorting incoming requests — build an MVP on your data and look at the numbers. If it pays off, we expand; if not, we stop, and everything done stays with you.
Contact
Describe what you want to automate — I’ll tell you whether AI fits, how long it will take and where to start.
Write it as it is, even if the task fits in two lines for now. Figuring out what’s really needed is already my job.
