Enterprise Sales Intelligence Platform

A natural-language business-intelligence platform letting non-technical users query enterprise sales data without writing SQL.

Business Problem

Sales information was distributed across 10+ connected tables in Oracle Autonomous Database. Business users depended on SQL developers and analysts for every report, and needed technical help interpreting the results.

Users

Non-technical sales and business teams who need answers from operational data directly, without a request-and-wait cycle through analysts.

Constraints

10+ connected Oracle tables Governed query execution Non-technical end users Low-latency target

Solution

I designed a Text-to-SQL system that converts natural-language business questions into SQL, executes it against approved Oracle schema, and returns the result as a plain-language explanation.

Architecture

User Question

→

Intent Interpretation

→

Schema Context Selection

→

SQL Generation

→

Validation & Execution

→

NL Explanation

Your Contribution

Designed

the Text-to-SQL architecture and schema context strategy.

Integrated

Oracle Autonomous Database across 10+ business tables.

Built

the SQL-generation workflow and backend API.

Added

query-validation controls and natural-language explanations.

Technology Stack

Python FastAPI SQL Oracle Autonomous Database LLM Engineering Prompt Engineering

Results

10+

tables queryable in plain language

~10s

average response latency

↓

manual SQL requests

Challenges & Trade-Offs

Flexibility vs. safety, accuracy vs. latency

A validation step checks generated SQL before execution — trading some flexibility for guardrails against unsafe or malformed queries, while keeping the full question-to-answer cycle within a ~10-second response target.

Confidentiality

Enterprise project. Database structures and real sales information are not shown publicly — this page describes the workflow, not the data.