APEX: Autonomous Platform for Enterprise eXcellence

Main data view with supported Data Cleaning steps to prepare for training stage
No-code, plug and play Training Lab, featuring training predictive model on user's data
Chat system, Exa for finding live trends and competitor pricing to analyze changes in internal data, LLM graph builder for visualization
Customizable dashboards to adapt to different use cases, users and scenarios

Inspiration

For Small and Medium Enterprises (SMEs), data isn't an asset — it's a bottleneck. While global corporations leverage elite data departments, smaller businesses often drown in stale spreadsheets, suffering from Decision Latency. We were inspired to build the "Great Equalizer": a tool that gives the underdog the brainpower of a Fortune 500 data department, turning raw data into immediate, strategic action.

What it does

Proactive Auditing: Automatically scans for anomalies, falling profits, and inventory risks (under 20 units) upon upload.
Market Intelligence: Triggers Web Research APIs to cross-reference internal drops with competitor pricing and global trends.
Predictive Lab: Auto-trains ML models (XGBoost/LightGBM) with professional data cleaning (KNN Imputation) to forecast future risks.
Autonomous Execution: Drafts supplier emails, restock lists, and sends critical Slack alerts to in-charge personnel for 30-second approvals.

How we built it

Backend: Built with FastAPI for high-performance asynchronous task handling.
Data Engine: Utilized DuckDB for lightning-fast local processing of tabular data.
The Brain: Implemented a ReAct (Reason + Act) agentic workflow using LangGraph to manage complex decision cycles.
Predictive Power: Integrated XGBoost and Scikit-learn for automated model training and data preservation.
Connectivity: Developed custom tools for Exa Web Search and Slack Block Kit integration to close the loop between insight and action.

Challenges we ran into

Closing the Loop: Moving beyond a chatbot that just "talks" to an agent that "acts" required complex state management to ensure the AI didn't hallucinate business actions.
Data Integrity: Handling "messy" SME data required implementing professional-grade preprocessing (like KNN imputation) so predictive models remained accurate even with missing inputs.
Decision Latency: Engineering the system to process over 2,500 rows of data, perform web research, and generate a strategy in under 30 seconds.

Accomplishments we're proud of

The "Lego" Dashboard: A fully customizable, drag-and-drop UI that adapts to different business roles (Sales, Logistics, Finance).
Zero-Prompt Awareness: An "Observer" logic where the bot proactively identifies a crisis — like falling profits — before the user even asks.

What we learned

Context is King: An AI is only as good as the playbook it follows. Learning to feed the agent company-specific constraints was a game-changer for accuracy.
Human-in-the-Loop: The most effective automation doesn't replace the human — it empowers them with 30-second approval workflows.

What's next for APEX: Autonomous Platform for Enterprise eXcellence

Deep Integration: Connecting directly to ERP systems like SAP and Odoo for real-time data streaming.
Multi-Agent Swarms: Developing specialized sub-agents for HR, Marketing, and Legal to expand the "Data Department" ecosystem.
Edge Deployment: Enhancing our local-first architecture to allow SMEs to run APEX on-premise for maximum data privacy.