The Problem
Enterprise ERP systems store massive amounts of critical business data — sales figures, inventory levels, order statuses — but accessing meaningful insights from this data traditionally requires SQL expertise or expensive BI tools. Business users and managers who need quick answers are forced to wait for data analysts to write queries, creating bottlenecks and slowing decision-making. There is no intuitive way for non-technical stakeholders to interact directly with their own business data.
Our Solution
We built a full-stack ERP Analytics Dashboard that combines a visual analytics layer with a conversational AI interface. Business users can either view pre-built interactive charts showing key ERP metrics, or simply type a question in plain English — such as “Which region had the highest sales in Q4?” — and receive an instant answer backed by live data from Snowflake. The system uses Groq-hosted LLaMA 3.3 70B to translate natural language into SQL, executes it against Snowflake in real time, and returns both the raw results and a human-readable summary.
Solution Architecture
User (Browser)
│
▼
Frontend (HTML + React + Recharts)
– Dashboard Tab: 6 pre-built charts
– Ask AI Tab: Chat interface with suggestions
│
▼
FastAPI Backend (Python)
– /sales/*, /inventory/*, /orders/* → chart data endpoints
– /chat → NL-to-SQL pipeline
│
├──► Groq API (LLaMA 3.3 70B)
│ – Step 1: Question → SQL generation
│ – Step 2: Results → Natural language answer
│
└──► Snowflake (Cloud Data Warehouse)
– ERP_DB.ERP_SCHEMA
– Tables: SALES, INVENTORY, ORDERS
Deliverables
- A fully functional ERP Analytics Dashboard with 6 interactive charts covering revenue trends, regional sales, top products, inventory levels, order status, and monthly order volume.
- A conversational AI chatbot interface that accepts natural language questions and returns SQL-backed answers from live Snowflake data.
- A FastAPI backend with 7 REST endpoints serving both chart data and the NL-to-SQL chat pipeline.
- A Snowflake database schema with seeded ERP sample data across 3 tables (SALES, INVENTORY, ORDERS).
- A ready-to-run local development setup with environment-based configuration.
Tech Stack
| Layer | Technology |
|---|---|
| Cloud Data Warehouse | Snowflake (AWS ap-southeast-7) |
| Backend Framework | FastAPI (Python) |
| LLM Provider | Groq (LLaMA 3.3 70B Versatile) |
| Frontend | React (CDN) + Recharts |
| Snowflake Connector | snowflake-connector-python |
| Server | Uvicorn |
| Environment Config | python-dotenv |
Business Impact
This solution directly addresses one of the most common pain points in enterprise data operations — the gap between data availability and data accessibility.
By enabling natural language querying on top of a cloud data warehouse, it eliminates the dependency on SQL-skilled analysts for routine business queries. A sales manager can instantly ask “Which product had the lowest units sold last quarter?” without raising a ticket or waiting for a report. This reduces query turnaround time from hours or days to under 10 seconds.
The architecture is production-scalable — Snowflake handles petabyte-scale data, Groq delivers sub-second LLM inference, and FastAPI supports high-concurrency API traffic. The same pattern can be applied across industries including retail, manufacturing, logistics, and finance wherever ERP data needs to be democratized for non-technical users.
For businesses, the key impacts are faster decision-making, reduced analyst overhead, lower barrier to data access, and a foundation for building more advanced AI-driven analytics products on top of existing data infrastructure.
Project Snapshots
Take at least 5 screenshots covering:
The Problem
Enterprise ERP systems store massive amounts of critical business data — sales figures, inventory levels, order statuses — but accessing meaningful insights from this data traditionally requires SQL expertise or expensive BI tools. Business users and managers who need quick answers are forced to wait for data analysts to write queries, creating bottlenecks and slowing decision-making. There is no intuitive way for non-technical stakeholders to interact directly with their own business data.
Our Solution
We built a full-stack ERP Analytics Dashboard that combines a visual analytics layer with a conversational AI interface. Business users can either view pre-built interactive charts showing key ERP metrics, or simply type a question in plain English — such as “Which region had the highest sales in Q4?” — and receive an instant answer backed by live data from Snowflake. The system uses Groq-hosted LLaMA 3.3 70B to translate natural language into SQL, executes it against Snowflake in real time, and returns both the raw results and a human-readable summary.
Solution Architecture
User (Browser)
│
▼
Frontend (HTML + React + Recharts)
– Dashboard Tab: 6 pre-built charts
– Ask AI Tab: Chat interface with suggestions
│
▼
FastAPI Backend (Python)
– /sales/*, /inventory/*, /orders/* → chart data endpoints
– /chat → NL-to-SQL pipeline
│
├──► Groq API (LLaMA 3.3 70B)
│ – Step 1: Question → SQL generation
│ – Step 2: Results → Natural language answer
│
└──► Snowflake (Cloud Data Warehouse)
– ERP_DB.ERP_SCHEMA
– Tables: SALES, INVENTORY, ORDERS
Deliverables
- A fully functional ERP Analytics Dashboard with 6 interactive charts covering revenue trends, regional sales, top products, inventory levels, order status, and monthly order volume.
- A conversational AI chatbot interface that accepts natural language questions and returns SQL-backed answers from live Snowflake data.
- A FastAPI backend with 7 REST endpoints serving both chart data and the NL-to-SQL chat pipeline.
- A Snowflake database schema with seeded ERP sample data across 3 tables (SALES, INVENTORY, ORDERS).
- A ready-to-run local development setup with environment-based configuration.
Tech Stack
| Layer | Technology |
|---|---|
| Cloud Data Warehouse | Snowflake (AWS ap-southeast-7) |
| Backend Framework | FastAPI (Python) |
| LLM Provider | Groq (LLaMA 3.3 70B Versatile) |
| Frontend | React (CDN) + Recharts |
| Snowflake Connector | snowflake-connector-python |
| Server | Uvicorn |
| Environment Config | python-dotenv |
Business Impact
This solution directly addresses one of the most common pain points in enterprise data operations — the gap between data availability and data accessibility.
By enabling natural language querying on top of a cloud data warehouse, it eliminates the dependency on SQL-skilled analysts for routine business queries. A sales manager can instantly ask “Which product had the lowest units sold last quarter?” without raising a ticket or waiting for a report. This reduces query turnaround time from hours or days to under 10 seconds.
The architecture is production-scalable — Snowflake handles petabyte-scale data, Groq delivers sub-second LLM inference, and FastAPI supports high-concurrency API traffic. The same pattern can be applied across industries including retail, manufacturing, logistics, and finance wherever ERP data needs to be democratized for non-technical users.
For businesses, the key impacts are faster decision-making, reduced analyst overhead, lower barrier to data access, and a foundation for building more advanced AI-driven analytics products on top of existing data infrastructure.
Project Snapshots
Take at least 5 screenshots covering:
The Problem
Enterprise ERP systems store massive amounts of critical business data — sales figures, inventory levels, order statuses — but accessing meaningful insights from this data traditionally requires SQL expertise or expensive BI tools. Business users and managers who need quick answers are forced to wait for data analysts to write queries, creating bottlenecks and slowing decision-making. There is no intuitive way for non-technical stakeholders to interact directly with their own business data.
Our Solution
We built a full-stack ERP Analytics Dashboard that combines a visual analytics layer with a conversational AI interface. Business users can either view pre-built interactive charts showing key ERP metrics, or simply type a question in plain English — such as “Which region had the highest sales in Q4?” — and receive an instant answer backed by live data from Snowflake. The system uses Groq-hosted LLaMA 3.3 70B to translate natural language into SQL, executes it against Snowflake in real time, and returns both the raw results and a human-readable summary.
Solution Architecture
User (Browser)
│
▼
Frontend (HTML + React + Recharts)
– Dashboard Tab: 6 pre-built charts
– Ask AI Tab: Chat interface with suggestions
│
▼
FastAPI Backend (Python)
– /sales/*, /inventory/*, /orders/* → chart data endpoints
– /chat → NL-to-SQL pipeline
│
├──► Groq API (LLaMA 3.3 70B)
│ – Step 1: Question → SQL generation
│ – Step 2: Results → Natural language answer
│
└──► Snowflake (Cloud Data Warehouse)
– ERP_DB.ERP_SCHEMA
– Tables: SALES, INVENTORY, ORDERS
Deliverables
- A fully functional ERP Analytics Dashboard with 6 interactive charts covering revenue trends, regional sales, top products, inventory levels, order status, and monthly order volume.
- A conversational AI chatbot interface that accepts natural language questions and returns SQL-backed answers from live Snowflake data.
- A FastAPI backend with 7 REST endpoints serving both chart data and the NL-to-SQL chat pipeline.
- A Snowflake database schema with seeded ERP sample data across 3 tables (SALES, INVENTORY, ORDERS).
- A ready-to-run local development setup with environment-based configuration.
Tech Stack
| Layer | Technology |
|---|---|
| Cloud Data Warehouse | Snowflake (AWS ap-southeast-7) |
| Backend Framework | FastAPI (Python) |
| LLM Provider | Groq (LLaMA 3.3 70B Versatile) |
| Frontend | React (CDN) + Recharts |
| Snowflake Connector | snowflake-connector-python |
| Server | Uvicorn |
| Environment Config | python-dotenv |
Business Impact
This solution directly addresses one of the most common pain points in enterprise data operations — the gap between data availability and data accessibility.
By enabling natural language querying on top of a cloud data warehouse, it eliminates the dependency on SQL-skilled analysts for routine business queries. A sales manager can instantly ask “Which product had the lowest units sold last quarter?” without raising a ticket or waiting for a report. This reduces query turnaround time from hours or days to under 10 seconds.
The architecture is production-scalable — Snowflake handles petabyte-scale data, Groq delivers sub-second LLM inference, and FastAPI supports high-concurrency API traffic. The same pattern can be applied across industries including retail, manufacturing, logistics, and finance wherever ERP data needs to be democratized for non-technical users.
For businesses, the key impacts are faster decision-making, reduced analyst overhead, lower barrier to data access, and a foundation for building more advanced AI-driven analytics products on top of existing data infrastructure.
Project Snapshots
Take at least 5 screenshots covering:
The Problem
Enterprise ERP systems store massive amounts of critical business data — sales figures, inventory levels, order statuses — but accessing meaningful insights from this data traditionally requires SQL expertise or expensive BI tools. Business users and managers who need quick answers are forced to wait for data analysts to write queries, creating bottlenecks and slowing decision-making. There is no intuitive way for non-technical stakeholders to interact directly with their own business data.
Our Solution
We built a full-stack ERP Analytics Dashboard that combines a visual analytics layer with a conversational AI interface. Business users can either view pre-built interactive charts showing key ERP metrics, or simply type a question in plain English — such as “Which region had the highest sales in Q4?” — and receive an instant answer backed by live data from Snowflake. The system uses Groq-hosted LLaMA 3.3 70B to translate natural language into SQL, executes it against Snowflake in real time, and returns both the raw results and a human-readable summary.
Solution Architecture
User (Browser)
│
▼
Frontend (HTML + React + Recharts)
– Dashboard Tab: 6 pre-built charts
– Ask AI Tab: Chat interface with suggestions
│
▼
FastAPI Backend (Python)
– /sales/*, /inventory/*, /orders/* → chart data endpoints
– /chat → NL-to-SQL pipeline
│
├──► Groq API (LLaMA 3.3 70B)
│ – Step 1: Question → SQL generation
│ – Step 2: Results → Natural language answer
│
└──► Snowflake (Cloud Data Warehouse)
– ERP_DB.ERP_SCHEMA
– Tables: SALES, INVENTORY, ORDERS
Deliverables
- A fully functional ERP Analytics Dashboard with 6 interactive charts covering revenue trends, regional sales, top products, inventory levels, order status, and monthly order volume.
- A conversational AI chatbot interface that accepts natural language questions and returns SQL-backed answers from live Snowflake data.
- A FastAPI backend with 7 REST endpoints serving both chart data and the NL-to-SQL chat pipeline.
- A Snowflake database schema with seeded ERP sample data across 3 tables (SALES, INVENTORY, ORDERS).
- A ready-to-run local development setup with environment-based configuration.
Tech Stack
| Layer | Technology |
|---|---|
| Cloud Data Warehouse | Snowflake (AWS ap-southeast-7) |
| Backend Framework | FastAPI (Python) |
| LLM Provider | Groq (LLaMA 3.3 70B Versatile) |
| Frontend | React (CDN) + Recharts |
| Snowflake Connector | snowflake-connector-python |
| Server | Uvicorn |
| Environment Config | python-dotenv |
Business Impact
This solution directly addresses one of the most common pain points in enterprise data operations — the gap between data availability and data accessibility.
By enabling natural language querying on top of a cloud data warehouse, it eliminates the dependency on SQL-skilled analysts for routine business queries. A sales manager can instantly ask “Which product had the lowest units sold last quarter?” without raising a ticket or waiting for a report. This reduces query turnaround time from hours or days to under 10 seconds.
The architecture is production-scalable — Snowflake handles petabyte-scale data, Groq delivers sub-second LLM inference, and FastAPI supports high-concurrency API traffic. The same pattern can be applied across industries including retail, manufacturing, logistics, and finance wherever ERP data needs to be democratized for non-technical users.
For businesses, the key impacts are faster decision-making, reduced analyst overhead, lower barrier to data access, and a foundation for building more advanced AI-driven analytics products on top of existing data infrastructure.
Project Snapshots
Take at least 5 screenshots covering:
- The full dashboard view with all 6 charts
- Revenue Trend + Sales by Region charts (zoomed in)
- Inventory Levels + Order Status charts
- The Ask AI tab with the chat interface open
- A live query being answered — e.g. “Which region had the highest sales?” showing the SQL + result table + natural language answer
Demo Video
