NetZero

Inspiration

Tracking and reporting Scope 1, 2, and 3 carbon emissions across complex supply chains is notoriously complex, highly manual, and prone to costly human error. We built NetZero to tackle climate accountability head-on. By leveraging Codex 5.6, we created an end-to-end intelligent agent system capable of parsing unstructured sustainability documents, executing complex GHG (Greenhouse Gas) protocol calculations on the fly, and recommending actionable decarbonization strategies.

What it does

NetZero acts as an autonomous sustainability copilot for businesses and ESG auditors:

• Automated Data Extraction: Upload invoices, energy bills, or supply chain spreadsheets, and Codex 5.6 automatically writes and executes custom parsing routines to extract tabular emission factors.
• GHG Protocol Math & Modeling: Automatically computes emissions equations using standardized formula conversions:

$$\text{Emissions } (t\text{CO}_2e) = \text{Activity Data} \times \text{Emission Factor} \times \text{GWP}$$

• Code-Driven Insights & Reporting: Generates customized Python data visualizations and actionable reduction roadmaps in seconds.

How we built it

• AI Engine & Code Generation: Powered by Codex 5.6 via the OpenAI API to dynamically write, validate, and execute Python scripts for complex mathematical conversions and document parsing.
• Agentic Workflows: Utilized Codex 5.6’s advanced reasoning and tool-use capabilities to build multi-step verification agents that check data accuracy before rendering final summaries.
• Backend: Built with Python (FastAPI) to safely execute generated code snippets in a sandboxed environment and manage data pipelines.
• Frontend: Responsive dashboard designed to display real-time emission analytics, audit trails, and downloadable compliance reports.

Challenges we ran into

• Format Heterogeneity: Utility bills and supplier disclosures come in wildly different formats. Prompting Codex 5.6 to handle extreme edge cases required robust error handling and structured JSON outputs.
• Deterministic Math Accuracy: Language models can occasionally make arithmetic slips. To guarantee 100% precision, we instructed Codex 5.6 to generate pure, executable Python functions for calculations rather than relying on standard token generation for math.

Accomplishments that we're proud of

• Seamlessly integrating Codex 5.6 to dynamically program data transformation pipelines on unstructured inputs with near-zero latency.
• Building a full-stack, enterprise-ready carbon accounting prototype within the hackathon timeframe.

What we learned

• Best practices for prompt engineering and structured code synthesis using Codex 5.6.
• Deepened our understanding of global ESG frameworks (GHG Protocol, CSRD) and how AI can drastically lower compliance overhead.

What's next for NetZero

• Expanding Codex 5.6 code execution capabilities to directly interface with enterprise ERP APIs (SAP, Salesforce, NetSuite).
• Predictive scenario modeling to simulate how renewable energy adoption impacts long-term corporate emissions targets.

Built With

• codex