About Client
The client is the world’s second-largest metals and mining corporation, operating globally across aluminium, diamonds, copper, gold and industrial minerals. Their trading teams rely on daily market intelligence from leading banks, but the high volume and diversity of reports made it difficult to extract the most relevant signals fast enough to inform trading decisions.
Objective
Traders faced information overload: hundreds of daily reports arriving from multiple banks meant it was no longer feasible to read and act on everything. The business needed an AI-powered, reusable framework capable of automating sentiment extraction and summarization from high-volume commodity market reports. The solution needed to process diverse sources, enable daily inference cycles, and serve as a modular foundation for future GenAI applications—reducing manual intervention to zero.
Primary Goals
- Automate extraction, classification and summarization of market views for aluminium and copper.
- Produce a clean, daily table for traders (Bank, View, Summary, Sentiment, File Link).
- Eliminate manual file copying/triggering and reduce time-to-insight.
- Build a reusable, CI/CD-enabled framework to accelerate future models and use cases.
Solution
Blazeclan designed and implemented a GenAI-powered NLP framework on AWS, purpose-built for high-volume, daily sentiment analysis.
Key solution highlights:
- AI-Powered Sentiment Analysis: Deployed LLM-inspired NLP/NLU models to classify and summarize aluminium and copper market insights from daily reports.
- Reusable AWS SageMaker Framework: Developed a modular and reusable infrastructure to support model development, validation, deployment, and monitoring for daily inference cycles.
- Automated DevOps Integration: Integrated with CI/CD pipelines using AWS DevOps tools and GitHub for automation, version control, and experiment tracking.
- Fully Automated Workflow: Designed an end-to-end pipeline covering Processing Job → Training Job → Prediction Job, enabling zero manual intervention.
- Future-Ready Architecture: Built the framework as a scalable, reusable foundation for future GenAI applications, minimizing future development timelines.
Outcome
- 0% Manual Effort: Fully automated ingestion, processing, and analysis.
- Reusable AI Framework: Accelerated deployment for future AI/ML initiatives.
- Faster Decision-Making: Near real-time sentiment and market view generation.
- Autonomous Operations: Scalable to handle large, complex datasets daily.
- Enhanced Market Intelligence: Structured summaries and sentiment tables for quick, actionable insights.
Tech Stack
- Amazon S3
- Amazon Sagemaker
- AWS Lambda
- NLP
- GitHub