You will get an AI analytics assistant built on your own data, plus a reporting portal
Project details
Your executives ask the same questions every week, and someone spends half a day pulling each answer.
I build a reporting portal with an AI assistant grounded in your own data — not a generic chatbot with your logo on it. Ask a question in plain English, it queries your data and returns the answer with the numbers behind it.
How it's built: your sources consolidated into a DuckDB analytics layer, Python analysis and Streamlit dashboards on top, and an LLM connected through a read-only query layer with a grounding document describing your schema and business logic. Delivered as a Docker container that runs on your VM, your cloud, or Streamlit Cloud — your data never leaves your infrastructure.
Safety is built in: read-only connection, an allowlist of tables the assistant can reach, and validation of every generated query before it runs.
Every tier includes an accuracy pass. You supply real questions, I test against them and document where it performs well and where it doesn't.
I run a production system like this at a company with 500+ locations, where it deflects 10 to 15 routine analytics requests a week.
Not sure this is the right starting point? Message me
I build a reporting portal with an AI assistant grounded in your own data — not a generic chatbot with your logo on it. Ask a question in plain English, it queries your data and returns the answer with the numbers behind it.
How it's built: your sources consolidated into a DuckDB analytics layer, Python analysis and Streamlit dashboards on top, and an LLM connected through a read-only query layer with a grounding document describing your schema and business logic. Delivered as a Docker container that runs on your VM, your cloud, or Streamlit Cloud — your data never leaves your infrastructure.
Safety is built in: read-only connection, an allowlist of tables the assistant can reach, and validation of every generated query before it runs.
Every tier includes an accuracy pass. You supply real questions, I test against them and document where it performs well and where it doesn't.
I run a production system like this at a company with 500+ locations, where it deflects 10 to 15 routine analytics requests a week.
Not sure this is the right starting point? Message me
AI Algorithms
Large Language Model, Transformer ModelAI Applications
AI Chatbot, Anomaly Detection, Conversational AI, Natural Language Generation, Natural Language Understanding, Time Series Analysis, Time Series ForecastingAI Development Language
PythonAI Tools
Azure OpenAI, StreamlitAI Models
ChatGPT, GPT-3, GPT-4, OpenAI CodexWhat's included
| Service Tiers |
Starter
$2,000
|
Standard
$4,500
|
Advanced
$9,500
|
|---|---|---|---|
| Delivery Time | 14 days | 28 days | 45 days |
Number of Revisions | 2 | 3 | 3 |
AI Model Integration | |||
Batch Normalization | - | - | - |
Database Integration | |||
Detailed Code Comments | - | ||
Image Upscaling | - | - | - |
MLOps | - | - | - |
Model Deployment | - | ||
Model Documentation | |||
Model Monitoring | - | - | - |
Model Testing & Optimization | |||
Model Tuning | - | - | - |
Natural Language Processing | |||
NLP Tokenization | - | - | - |
Pre-Training | - | - | - |
Prompt Engineering | |||
Setup File | |||
Source Code | - |
Optional add-ons
You can add these on the next page.
Additional Revision
+$350Frequently asked questions
About Shemal
BI & Analytics Leader | Power BI, Fabric, SQL & AI Automation
Mississauga, Canada - 1:56 am local time
Three times now I've joined a company where the data function was tiny or nonexistent and built it into the thing leadership runs the business on. Today I lead BI and analytics for a gaming portfolio covering 500+ locations and an iGaming arm processing over $800M in annual bet volume, reporting directly to the CEO, CFO, and CTO.
WHAT I BUILD FOR CLIENTS
Executive Power BI & Microsoft Fabric environments — semantic models, service deployment, Copilot-enhanced dashboards, board-ready reporting
Revenue forecasting — ensemble time-series (Holt-Winters, blended-growth, YTD-anchored), backtested with MAPE and confidence bands, plus variance and bridge analysis
AI automation for BI — retrieval- and tool-augmented assistants grounded in your own data, and custom MCP servers connecting LLMs to SQL Server, Oracle, and Jira
Statistical modeling that moves revenue — Taguchi DOE, multiple regression, K-means clustering, Monte Carlo simulation
Data consolidation — messy Power BI, SQL Server, Oracle, and Excel sprawl unified into one warehouse layer that everyone actually trusts
WHY THE STATISTICS BACKGROUND MATTERS
My PhD and a granted patent came out of optimizing concrete using experimental design. The same rigour that tunes a material tunes a game economy or a revenue forecast — I just point it at different problems now. Most BI freelancers can build you a dashboard. Fewer can tell you whether the number on it is significant.
A FEW RESULTS
Game-economy and player modeling that lifted Gross Gaming Revenue 19% and player retention 11%
Custom MCP servers automating reporting workflows — ~20 analyst hours reclaimed per week
An AI analytics assistant deflecting 10–15 routine reporting requests weekly, about an hour of team time a day
Sales and QA dashboards at a manufacturer contributing to a 14% rise in quarterly revenue and ~10 hours per week in process savings
HOW I WORK
I start by asking what decision you're trying to make, not what chart you want. Then I scope it, build it, and hand it over documented so your team isn't dependent on me forever. PMP-certified, so scope, timeline, and status reporting are handled without you chasing me.
You'll find a live retrieval-augmented AI analytics assistant, forecasting work, and full case studies in my portfolio below.
Based in the Greater Toronto Area, Canadian Permanent Resident, and comfortable working across North American and European hours.
If you're building (or rebuilding) an analytics function, message me with what's broken and I'll tell you honestly whether I'm the right fit.
Steps for completing your project
After purchasing the project, send requirements so Shemal can start the project.
Delivery time starts when Shemal receives requirements from you.
Shemal works on your project following the steps below.
Revisions may occur after the delivery date.
Scoping and question capture
I map your data sources, agree which tables are in scope, and collect the real questions the assistant needs to answer. Scope is fixed here in writing before any build starts.
Warehouse layer build
Your sources consolidated into a DuckDB analytics layer so the dashboards and the assistant read the same numbers. Refresh schedule agreed and configured.