AMD Financial Stock Intelligence Agent
Solution overview
In this lab you will have access to a deployment of an AMD blueprint for a Financial Stock Intelligence Agent and learn about the full stack that it is optimized for, starting with the underlying AMD hardware environment, then moving up the AMD AI stack, and finally landing with the full stack software solution. The various layers of the architecture will be clearly explained allowing you to investigate the benefits that each component brings to power the full solution.
Financial Stock Intelligence Agent
The Financial Stock Intelligence Agent leverages AI-powered analysis to help financial analysts quickly uncover actionable insights from market data. By combining real-time stock prices from Yahoo Finance with a suite of technical indicators — including Simple Moving Average (SMA), Relative Strength Index (RSI), and momentum analysis — it produces natural language commentary and visualizations that surface trends, signals, and risks across any publicly traded equity. Analysts can enter a ticker symbol, set a date range, and receive a comprehensive report in seconds, blending quantitative indicators with LLM-generated interpretation. This enables teams to move from raw data to informed perspective without requiring deep technical expertise or proprietary tooling.
AMD Instinct Platform
Unlike solutions that rely on a single integrated appliance, the Financial Stock Intelligence Agent is deployed directly on Kubernetes — giving organizations the flexibility to run it on bare metal, on-premises clusters, or in a cloud environment of their choosing. The foundation of this deployment is the AMD Instinct MI250, AMD's data center GPU engineered for HPC and AI workloads. At the heart of the MI250 is AMD's CDNA2 architecture, with 128GB of HBM2e memory per GPU split across two dies — providing substantially more GPU memory than many alternative platforms and enabling large models to run fully in-memory without quantization trade-offs.
This platform raises the same operational questions any IT organization must answer when deploying AI:
- Where is our data processed and stored?
- How does this integrate with our existing infrastructure?
- How do we maintain performance and control at scale?
Deploying on Kubernetes with AMD Instinct addresses these questions by giving teams direct control over their environment. The solution is packaged as an OCI-compliant Helm chart, making deployment, version control, and teardown straightforward. No proprietary management layer is required, just a Kubernetes cluster with AMD Instinct GPUs and standard tooling.
AMD AI Stack
The MI250's 128GB of HBM2e memory unlocks model sizes that constrained platforms cannot serve without performance penalties. For this use case, we have deployed Gemma 3 27B as the primary inference model. The model is served via AIM (AMD Inference Microservice), AMD's containerized inference serving solution designed for secure, reliable deployment of high-performance AI models on AMD hardware. AIM provides an OpenAI-compatible API surface, making it straightforward to integrate into existing application architectures and swap models as needs evolve. This solution relies on:
- Gemma 3 27B via AIM
- AMD ROCm — AMD's open-source GPU compute platform powering model execution on the MI250
To orchestrate the financial analysis pipeline, the blueprint uses LangChain as the foundational AI framework. LangChain is a widely adopted open-source framework for building context-aware reasoning applications, enabling the agent to chain together data retrieval, indicator computation, and LLM inference into a coherent analytical workflow. LangChain's extensive ecosystem of integrations allows rapid connection of real-time market data, technical analysis logic, and the AIM-hosted LLM into a production-ready pipeline.
Full Stack AI Solution
The last layer of this lab focuses on the full stack solution that powers the Financial Stock Intelligence Agent. Built using Python with a Gradio web interface, the application serves as an example of how a complete, end-to-end AI solution can be assembled from open-source components on open hardware. Gradio provides an accessible, browser-based UI that requires no front-end development expertise to customize, making it well suited for rapid prototyping and stakeholder demonstration alike. The analytical backend handles market data retrieval, computes technical indicators, and coordinates with the AIM LLM to produce the final output, all within a single containerized service.
At WWT, we understand the need to build solutions that seamlessly fit into our clients' organizational landscape. Some portions of an AI application will always be new, but the other technical decisions don't have to be. By building the Financial Stock Intelligence Agent on standard Kubernetes primitives, open-source AI frameworks, and AMD's open software stack, this solution avoids proprietary lock-in and is positioned to scale from a single ATC demonstration to enterprise deployment.