Nikolay
Romanovskiy

Average Review

0.00

Reputation

Finished gigs

Finished jobs

Website

github.com/TheBad007

Country

Not specified

Specialization

Python Developer (AI Agents & Automation)

Hourly rate

$25/hr

Preferred payment options

USDT

About me

Python Developer: AI agents, LLM integrations, Telegram bots, parsers, automation. I build AI support and sales agents (OpenAI/Claude API: tool calling, RAG knowledge bases that answer from your documents with a link to the source), Telegram bots and Mini Apps (aiogram 3, FastAPI, PostgreSQL), parsers and data collection pipelines, n8n/Make automations — CRM, Telegram, spreadsheets, AI processing. Backend: FastAPI, PostgreSQL, Docker, VPS deployment. How I work: fixed price agreed after clarifying questions — it never changes mid-project. Work is delivered in milestones with intermediate demos; the code and repository stay with you. 50% upfront, and after delivery you get two weeks of free bug fixes. Message me in chat — describe your task and I'll quote a price and deadline within a day.

TG: @Nikolay_myname

Work experience

Freelance

February 2019 – Current time

Job title

Python Developer (AI Agents & Automation)

Work experience & achievements

ython Developer: AI agents, LLM integrations, Telegram bots, parsers, automation. I build AI support and sales agents (OpenAI/Claude API: tool calling, RAG knowledge bases that answer from your documents with a link to the source), Telegram bots and Mini Apps (aiogram 3, FastAPI, PostgreSQL), parsers and data collection pipelines, n8n/Make automations — CRM, Telegram, spreadsheets, AI processing. Backend: FastAPI, PostgreSQL, Docker, VPS deployment. How I work: fixed price agreed after clarifying questions — it never changes mid-project. Work is delivered in milestones with intermediate demos; the code and repository stay with you. 50% upfront, and after delivery you get two weeks of free bug fixes. Message me in chat — describe your task and I'll quote a price and deadline within a day.

Education

Not specified

No gigs added yet

AI Support Agent: Knowledge Base + Human Operator Escalation

A support agent built in Python + LLM API: it classifies each incoming ticket on its own (category, priority), answers customers strictly from a knowledge base, and hands the conversation over to a live operator whenever it's unsure or something fails. Key design decisions: - hallucination protection: every LLM response is validated with pydantic schemas; an invalid response = escalation — the customer never sees garbage; - fail-closed: an LLM outage, timeout, or empty knowledge base always leads to a human operator; - escalation queue with priorities, dialog memory, rate limiting to protect the API budget; - metrics: auto-answer (deflection) rate, escalation reasons, p50/p95 latency; - 32 automated tests. Stack: Python 3.11, OpenAI/Claude (switchable via config, with a mock mode that requires no API keys), FastAPI, aiogram 3. Connects to Telegram or a website via HTTP API.

Backtest Engine Audit & Development for Trading Bots

Why most open-source trading bots lose money under independent verification: sign errors in funding payments, lookahead leaks, and no accounting for liquidity. What I verify in an engine (invariants, each one covered by an automated test): capital conservation recomputed from the trade log, funding signs (long pays when the rate is positive), signals executed strictly at the next bar's open, fail-closed on data gaps, capacity clips on orders, and independent PnL recomputation using a second method. Demo is public: engine + 14 automated tests, runs with a single command, no external APIs. The methodology was battle-tested on Binance/Deribit data from 2020–2026. Format: an audit of your backtest with a report, or building an engine from scratch to this standard.

Telegram Bots & Mini Apps with AI

Development of Telegram bots and Mini Apps: from simple scenario bots to AI agents with dialog memory that answer from your own documents. - aiogram 3, dialog states, webhook/polling - LLM integration (OpenAI/Claude): memory, tool calling - payments: Telegram Stars, crypto - deployment: Docker + VPS, logging Typical timelines: simple bot 3–5 days, AI agent 7–14 days, Mini App — from 3 weeks.

NikolayRomanovskiy