Oleksii Patsurkovskyi
Offer Oleksii work on your next project.
Private message
Offer work
Odessa,
Ukraine
3 days 1 hour back
available for hire
age 29 years
on the service 10 years
Rating
Successful projects
Average rating
CV
I build web products and automation solutions that actually work — not just “look done”.
- React / Next.js / Node.js
- AI integrations and content generation
- automation and scripting
- MVP development and fast product launches
I have a product mindset → I focus not only on code, but on results, UX, and business logic.
Experience includes:
- CRM systems (development + implementation)
- e-commerce (funnels, processes, automation)
- AI-driven projects (content, architecture, development acceleration)
If you have an idea or a task — I can help turn it into a working solution quickly.
Skills and abilities
Design & art
Promotion
Portfolio
-
550 USD Automation of financial accounting: Jobber → Google Sheet via N8N
Bot Development
A comprehensive and fault-tolerant automation system has been implemented for the cleaning business. The main task is to set up automatic data transfer from the CRM Jobber to Google Sheets using the n8n platform for accurate calculation of payments to cleaning teams (percentage and hourly model).
What has been done (Technical implementation details):
… - Sectional data model ("Constructor"): The visit row in the table is divided into independent zones (Visit, Invoice, Payment, Calculation). Each n8n flow updates only its fields, which eliminates conflicts during parallel processing of webhooks.
- Dynamic addressing (Header-based): n8n reads column numbers by keys in the first row of the table. The manager can freely swap columns — the integration will not break.
- Idempotency and duplicate protection: Unique keys event_id are generated, and atomic upsert mechanics are implemented through n8n Data Table. Repeated webhooks are automatically filtered out, preventing duplicates.
- Flexible financial distribution:
- An algorithm has been written to determine the Cash Taker among several teams on the visit based on their type and ID.
- Logic for splitting multi-visit invoices has been implemented (if one invoice is issued for several cleanings, amounts are divided equally).
- Manual Editing mode (Manual Lock): A data protection mechanism has been created. If the manager checks the manual correction box on any visit row, the automation completely freezes updates to this object to avoid overwriting manual edits.
- Fail-safe and monitoring:
- An error queue — Dead Letter Queue (Failed_Events) has been integrated directly into Google Sheets for convenient analysis by the manager.
- An Exponential Backoff retry policy has been set up to bypass API limits (429, 5xx). A system of 24 custom Telegram alerts (divided by branches: Info, Warnings, Errors) has been created for instant notification of discrepancies in amounts, absence of teams in directories, or API errors.
- Load optimization: The calculation of final payments (V), cleaning amounts (N), and discrepancies (R) has been completely offloaded to Google Sheets formulas within the row, reducing the number of API requests and speeding up system performance.
Technology stack:
- n8n (Workflow creation, n8n Data Table for logging and deduplication)
- Jobber API (GraphQL, Webhooks, Event verification via HMAC-SHA256)
- Google Sheets API (Batch updates, working with formulas)
- JavaScript / Node.js (Regular expressions, parsing line items, and data filtering logic within n8n nodes)
- Telegram Bot API (Routing notifications by topics)
Business result:
The client received a fully autonomous financial showcase. Manual work by the accountant/manager has been minimized — the system itself collects data on completed work, calculates net income after parking and taxes, divides tips, and calculates salaries for cleaners. All non-standard cases or financial discrepancies are highlighted via Telegram, ensuring 100% control over finances.
-
790 USD Backtesting algorithmic trading strategies
Python
Development of a testing system for trading algorithms on large arrays of historical market data. The feature was that strategies could have hundreds of thousands of different parameter combinations.
It was possible to implement backtesting of a million parameter combinations over a 2-year history in 2 hours.
… Tasks I solved:
Big data processing: Organized work with massive volumes of historical market data.
Mathematical calculations: Implemented complex logic, including matrix calculations and vectorization of operations for maximum processing speed.
Performance optimization: Used the Numba library for JIT compilation and eliminating performance bottlenecks in the system core.
Key skills: Python, Pandas, NumPy, Numba, Data Engineering, Algorithmic Trading, Matrix calculations, Vectorization.
-
750 USD Full-stack SaaS service for language learning with AI dictionaries
Javascript and Typescript
Creation from scratch of a web service (Next.js, Zustand, Supabase) for personalized language learning. Solving the problem of classic applications — lack of flexibility in choosing dictionaries and rigid binding to language pairs.
Tasks I addressed:
… Architecture design: Developed a universal word storage system that allows studying any combination of languages without duplicating logic in the code.
AI solutions integration: Implemented the generation of thematic dictionaries. The user specifies any narrow topic (for example, "medical terminology"), and AI generates a selection with translations and examples.
Content automation: Automated the localization process and database filling — over 8,000 words translated and added with the help of AI.
Results: Progressed from idea (CustDev) to a fully functioning and scalable SaaS product in one and a half months.
Key skills: Product Development, Next.js, Supabase, AI Integration, Prompt Engineering, Multilingual architecture.
-
2600 USD Development of a B2B CRM system for warehouses and e-commerce
Web Programming
Creation from scratch and full support of a custom CRM system designed for automating business processes of warehouses and online stores.
Tasks I addressed:
… Design and strategy: Conducted market research, gathered feedback from potential users, and developed a step-by-step product roadmap.
Development management: Wrote detailed technical requirements (specifications) for all system functions. Managed the full product lifecycle — from concept to successful on-time release.
Implementation and onboarding: Personally conducted consultations and training for clients. Assisted in integrating the CRM into their current realities and building new, more efficient business processes.
Support and development: Ensured uninterrupted operation and development of the system's functionality for almost 3 years based on product metrics and feedback.
Business results:
-30% operational load: Thanks to the automation of routine tasks and effective process structuring within the system, clients were able to reduce the workload on their employees by one third.
High level of loyalty: Clients specifically noted the quality of communication, structured approach to staff training, and deep immersion in their business details.
Key skills in the project: Product Management, Market Research, Roadmap Development, Client Communication, writing specifications, business process optimization.
-
280 USD Full-Stack dashboard for Telegram analytics (React, Node.js)
Javascript and Typescript
Full-stack application (SPA) for Telegram channel analytics. The system automatically detects abnormal spikes in audience interest and finds viral publications at early stages, using the relative metric Share Rate (repost rate).
A custom parser based on MTProto API (GramJS) has been developed, which collects message history on behalf of the user, bypassing the limitations of the standard Bot API. The core of the system calculates the baseline norm (Median) for each channel using a sliding data window (from T-8 to T-1 days), strictly filtering out informational noise and false anomalies.
… Key features:
Smart virality math: The algorithm compares fresh posts not by the number of views, but by the deviation of the repost percentage from the historical norm of a specific channel.
Safe parsing (Anti-Flood): A complex system for bypassing Telegram API blocks has been implemented (floating Jitter delays, pauses during pagination), simulating the behavior of a live person.
Lazy Media Download: To optimize disk space and reduce network load, media files are downloaded and cached by the backend only for confirmed viral posts.
Fault-tolerant database: Using SQLite in asynchronous logging mode (WAL) allows hundreds of records to be written in the background simultaneously and serve data to the frontend without interface freezes.
Autonomous authorization: Native UI input for the Telegram confirmation code directly in the React interface with secure session storage in the database.
Technology stack:
Frontend: React.js, Vite, Tailwind CSS (Dark Mode, Responsive Grid/Table layouts).
Backend: Node.js, Express.js.
API & Data: GramJS (Telegram MTProto Client), better-sqlite3 (WAL mode).
Activity
| Latest proposals 2 | Budget | Added | Deadlines | Proposal | |
|---|---|---|---|---|---|
|
Development of the front end for the CRM system (Frontend, React)
564 USD
|
|||||
|
TradingView indicator
582 USD
|
