INTERLINKED GLOBAL

Strategy, Simulated. Foresight, Quantified.

Comment

Inspiration

Strategic decision-making is often flawed in organisations. Leaders make key decisions without fully grasping the broader impact on their company’s structure. This leads to overlooked bottlenecks, hidden dependencies, and unexpected consequences. Traditional strategic tools are often just static documents that don’t adapt to changing internal dynamics or shifts in the market. That's why I build the "Ask it anything about your Organisation" strategic graph-aware agent...bridging the gap between people, processes and departments.

Our motivation came from the idea of a graph-aware AI that understands an organization's real-time context. By combining internal structure with external intelligence, this AI could offer strategic insights tailored to your specific environment. Imagine an AI that actually understands your company’s relationships, workflows, and limitations instead of offering generic advice.

What it does

INTERLINKED is a graph-aware AI strategy assistant that redefines how organisations make strategic decisions. When you ask something like, "How can we reduce engineering delivery time by 30 percent?", the system:

  1. Extracts live organizational context from your Neo4j graph database
  2. Searches external intelligence via Exa.ai for industry benchmarks and best practices
  3. Generates structured strategic analysis using our RL fine-tuned model
  4. Provides quantified recommendations with clear trade-offs and action plans
  5. Visualise the impact on your organization's graph in real time.

The result is a strategy that fits your organisation, not one-size-fits-all suggestions.

How we built it

Frontend: Next.js with Tailwind CSS, featuring dark mode and responsive design

Backend: FastAPI with real-time graph context extraction AI Pipeline: RL fine-tuned Qwen2.5-3B model on SageMaker Custom reward function optimizing for strategic framework fidelity Bedrock Claude 3 Haiku as fallback for reliability

Data sources include:

Neo4j for internal organizational context Exa.ai for external market data Structured XML outputs for consistent parsing and display

Our key innovation is the real-time integration of graph context into AI strategy formulation. This allows the system to understand your real structure, limitations, and processes.

Challenges we ran into

Model deployment was tricky. The 3B model hit memory limits on SageMaker. We solved it using CPU offloading and larger GPU instances There were compatibility issues with LoRA adapters due to version mismatches. We fixed this by creating lean bundles with only the necessary components Injecting Neo4j data in real time required precise engineering to keep responses relevant Building for production reliability meant setting up strong error handling, fallback processes, and root-cause tracking Time constraints were tight. We had to turn a research-level idea into a working MVP within days while keeping it usable and explainable

Accomplishments that we're proud of

Built a working graph-aware AI system in under one week Optimized a 3B model with a custom reward function focused on strategic reasoning Created the first real-time graph context injection system Achieved production-level reliability with solid error handling and backups Designed an intuitive interface that turns complex data into visual insights Integrated multiple AWS services into a streamlined system Proved the concept with real demos showing measurable trade-offs and impact

What we learned

Graph context is powerful. AI becomes far more useful when it understands your internal structure Reinforcement learning works for strategy. Custom reward functions help balance clarity, accuracy, and brevity Combining internal graphs with external data gives insights that neither can deliver on their own Building production systems is tough. What works in notebooks often fails in real applications. Error handling and reliability matter Structured outputs like XML improve consistency and help systems scale to production use Small, motivated teams can build incredible things fast when the vision is clear

What's next for INTERLINKED GLOBAL

Next 3 months:

  • Work with enterprise clients using their existing organizational data
  • Expand the strategic framework to cover more decision areas
  • Add integrations with tools like Slack and Microsoft Teams

6 to 12 months:

  • Launch a Strategy Tutor Mode that explains the frameworks behind each recommendation
  • Enable market simulations between multiple companies to explore strategic outcomes

Long-term:

  • Become the strategic foresight engine of the INTERLINKED GLOBAL platform
  • Start a pilot program with SMEs and universities across ASEAN for executive education
  • Add real-time team collaboration features for group-based strategy development

Built With

Share this project:

Updates