Sustainability and ESG teams were burdened with manually reviewing unstructured documents—PDFs, scanned reports, tables, and images—to analyze Scope 1, 2, and 3 emissions, identify compliance gaps, and benchmark peers. The process was inconsistent, slow, and lacked scalability.

Develop an AI-powered ESG Co-Pilot that could intelligently process varied document formats, extract critical ESG metrics using LLMs and NER models, support advanced semantic search, and deliver real-time, contextual analysis through a developer-friendly API interface.

• Implemented OCR and text extraction to ingest reports
• Used Named Entity Recognition (NER) and  LLMs (Llama-3, GPT-4o) for metric extraction (Scope 1/2/3, water, energy, waste)
• Stored structured metadata in MongoDB; stored vector embeddings in Qdrant for semantic search
• Designed query classification and routing to SQL, NoSQL, or Vector DB based on intent
• Built context-aware response generation (disclosures, benchmarks, gap analyses)
• Enabled real-time access through API endpoints like /agent-v3, /peer-benchmarking, /esg-disclosures

• Reduced ESG analysis time by over 70%
• Instant compliance gap detection and peer benchmarking
• Scalable, production-grade architecture using FastAPI, Redis, Qdrant, MongoDB, LangChain/Graph
• Became the core engine powering BCV’s AI-based Carbon & ESG Intelligence Platform

The solution can be deployed on Akamai’s edge-native infrastructure, enabling secure, low-latency API access for real-time ESG insights. Akamai Cloud’s globally distributed platform supports containerized deployment of the FastAPI stack, integrates with MongoDB and Qdrant instances, and ensures secure access through WAF, token auth, and data encryption. Edge computing can further optimize semantic query performance and compliance validation close to end-users.