Browse categories
Explore
Fiverr Pro
English
$
USD
I will integrate openai, claude or gemini into your app
Dmytro L
About this gig
The gap that kills AI features is not the prompt. It is everything around it: rate limits, timeouts, cost, and what your app does when the model returns nonsense.
WHAT I DO
Add an AI feature to the app you already have, or build the whole slice including UI and deploy. Summarization, extraction, classification, search, chat, structured output from messy input.
WHAT PRODUCTION MEANS HERE
Streaming, so users are not watching a spinner. Caching and model choice, so the bill stays sane. Retries and backoff on 429s and timeouts. Validation before anything reaches your database. Tests that cover the failure paths, wired into CI.
RECENT WORK
An AI content and translation engine for a SaaS I led as CTO: 4x more leads, site health from 11% to 100%. A crypto platform where smart order matching cut deal time from about 7 minutes to 1-2.
STACK
Python (FastAPI, Django), Node.js, React and Next.js in TypeScript. OpenAI, Claude, Gemini. Postgres, pgvector, Redis, Docker, AWS. pytest, mypy, GitHub Actions.
Send me your repo or a short description of your stack and I will tell you where the feature fits and which model I would use.
Get to know Dmytro L
Dmytro L
CTO level AI Engineer for RAG Chatbots AI Agents and LLM Integration
- FromUkraine
- Member sinceJul 2026
Languages
Ukrainian, Russian, English
I build production AI systems, not prototypes. 8+ years full-stack, last two roles as CTO leading projects end to end.
Recent work: a support RAG chatbot that handles around 80% of customer cases and escalates the rest. Agentic workflows for a medical app with guardrails and structured outputs. An AI content engine that grew a SaaS's leads 4x.
Stack: Python (FastAPI, Django), Node.js, React/Next.js. OpenAI, Claude, Gemini. LangGraph, OpenAI Agents SDK, CrewAI, n8n. pgvector, Pinecone. Docker, Kubernetes (LF certified), AWS.
Clean code, tests and deployment every time. Kyiv, UTC+3.
Other AI Development Services I Offer
FAQ
Can you work in my existing codebase?
Yes, that is the normal case. Send repo access or a zip and I work in your conventions, on a branch, with a PR you review. I do not rewrite what already works. If something in the repo will block the feature, I tell you before we start, not after.
Which model should I use and what will it cost to run?
Depends on the job. Classification and extraction usually run fine on a small cheap model; long reasoning needs a bigger one. I benchmark two or three on your real data, show quality and cost side by side, and build it so you can swap models later.
How do you handle rate limits and failures?
Timeouts on every call, exponential backoff on 429 and 5xx, a queue when volume spikes, and caching so repeated inputs never hit the API twice. If the model is down or returns something invalid, the app degrades gracefully instead of erroring at your user.
Do you write tests?
Always. Unit tests on the logic, contract tests at the model boundary with recorded responses so the suite does not burn tokens, and failure-path tests for timeouts and bad output. It all runs in GitHub Actions with ruff and mypy as gates.
Can you also build the UI?
Yes. React and Next.js in TypeScript, with Tailwind and shadcn if you like. Streamed responses, proper loading and error states, and the small details that make an AI feature feel fast. That is what the Premium package covers.
