You will get an AI Agent that chats with your Spreadsheet, 15k rows no hallucinations
Rising Talent
Project details
Spreadsheets hold your data, but getting answers out of them means formulas, filters, or pivot tables. This connects your Google Sheets to Claude or ChatGPT so you (or your team) just ask — "what's the total for Q2 by region?" — and get the real number back, computed from your data, not guessed by the model.
The difference from a normal "AI + Sheets" bot: answers are deterministic. The model writes a SQL query, the database returns actual rows, and the numbers come from SQL — so there are no hallucinated figures, even across multiple large spreadsheets.
Why it works for you
Ask in plain English — no formulas, no SQL, no training needed.
Grounded, accurate answers — figures are computed by the database, never invented.
Handles scale — multiple spreadsheets, 15,000+ rows, routed automatically to the right source.
Runs locally — deploy on your own machine/infra so your data stays with you.
Uses the tools you already have — works as an MCP server inside both Claude and ChatGPT.
What's included
A local MCP server wired to PostgreSQL, NL to SQL agent connected to your Google Sheets.
Automatic routing across sheets/sources.
Short Loom walkthrough + written instructions
The difference from a normal "AI + Sheets" bot: answers are deterministic. The model writes a SQL query, the database returns actual rows, and the numbers come from SQL — so there are no hallucinated figures, even across multiple large spreadsheets.
Why it works for you
Ask in plain English — no formulas, no SQL, no training needed.
Grounded, accurate answers — figures are computed by the database, never invented.
Handles scale — multiple spreadsheets, 15,000+ rows, routed automatically to the right source.
Runs locally — deploy on your own machine/infra so your data stays with you.
Uses the tools you already have — works as an MCP server inside both Claude and ChatGPT.
What's included
A local MCP server wired to PostgreSQL, NL to SQL agent connected to your Google Sheets.
Automatic routing across sheets/sources.
Short Loom walkthrough + written instructions
AI Development Type
Knowledge RepresentationAI Development Language
PythonWhat's included $400
These options are included with the project scope.
$400
- Delivery Time 3 days
- Number of Revisions 0
2 reviews
(2)
(0)
(0)
(0)
(0)
This project doesn't have any reviews.
KM
Katsiaryna M.
Apr 22, 2026
AI Developer for RAG / Knowledge Bases / Agentic Systems
Kyryl did an excellent job improving our internal AI platform built on LlamaIndex, LangChain, vector databases, and a Python/FastAPI backend. He enhanced the RAG pipeline, fixed retrieval inconsistencies, and noticeably improved chatbot logic and response quality. Highly recommend for any RAG, knowledge‑base, or agentic workflow development.
ID
Ivans D.
Apr 3, 2026
AI Engineer for internal AI system (RAG + LLM, real data)
Kyryl did an excellent job improving our internal AI system built on Python, OpenAI, and a vector database. He quickly fixed retrieval issues, reduced hallucinations, and significantly boosted answer quality across our Notion and Google Drive data sources. The pipeline is now far more stable, accurate, and production‑ready. Strongly recommend for any RAG/LLM engineering work.
About Kyryl
AI Engineer | RAG & LLM Agents | Production-Grade AI | Evaluations
100%
Job Success
Kyiv, Ukraine - 5:17 pm local time
Most AI contractors deliver demos. I deliver production systems. Before you commit, I'll run a live demo on your actual data — a working RAG pipeline, a trend-intelligence agent, or a Reddit analytics system — so you see real output before a single dollar changes hands. My stack goes from LangChain, LlamaIndex, and LangGraph all the way to custom LLM Ops, cost-optimized inference, and security layers (RBAC, guardrails, compliance). I've shipped multi-LLM agent systems combining Claude, Gemini, GPT, and open-source models — and I own the architecture end to end.
Here's what I build:
✅ RAG Systems — semantic + hybrid + BM25 retrieval, reranking, evaluation, and monitoring. Accurate answers, no hallucinations.
✅ LLM Agents & Agentic Workflows — autonomous pipelines that replace manual ops and remove burden from your team.
✅ Evaluations — labeled golden datasets, baseline benchmarking, and LLM-as-judge scoring.
✅ LLM Orchestration — GPT, Claude, Gemini, Qwen, Kimi, and open-source models, cost-optimized.
✅ AI Automations — custom Python pipelines that save measurable time.
✅ Backend & Infra — FastAPI, Node.js, pgvector, Pinecone, Weaviate, AWS, GCP.
Request a demo before ordering:
✨ RAG demo — production retrieval pipeline with hybrid search, reranking, and monitoring
✨ Reddit analytics demo — topic clustering, sentiment, trend detection, influencer mapping
✨ TikTok trend-finder demo — real-time trending video discovery with embeddings and ranking logic
"We thought we needed AI. What we actually needed was him. He rebuilt our entire approach — cut costs, improved accuracy, and delivered something our competitors still can't replicate." — Head of Growth, B2C SaaS
Selected builds: Computer vision models powering a 1M+ download iOS app. Revenue forecasting platform for $50M+ ARR SaaS products. Enterprise-grade RAG knowledge platform with real-time data and security layers. TikTok trend-intelligence agent using LangGraph and multi-LLM NLP pipelines.
Send me a message with your idea, your current bottleneck, or just the word 'AI' — I'll tell you exactly how to build it right, or why it hasn't been built right yet. I respond within a few hours.
Steps for completing your project
After purchasing the project, send requirements so Kyryl can start the project.
Delivery time starts when Kyryl receives requirements from you.
Kyryl works on your project following the steps below.
Revisions may occur after the delivery date.
Your order and preferences
Tell us if the deployment is local or in the cloud, any restrictions we should now. Let's char first
We send you an installer + instructions
You get the package, install and use. We are available for live support