Company Memory Guardian

Built by Ipsita Choudhury | Solo Builder

Inspiration

When experienced engineers leave, companies often lose the reasoning behind important decisions. The code remains, but the context, trade-offs, and lessons learned disappear. I wanted to build a company second brain that preserves this knowledge and helps teams make better decisions.

What it does

Company Memory Guardian turns engineering knowledge into actionable decision intelligence.

• Company Brain: AI-powered answers with citations to original GitHub sources.
• Decision Intelligence: Understand historical decisions, technical trade-offs, downstream risks, and potential ROI.
• Human Review: Simulated approval workflow for evaluating proposed engineering changes.
• Knowledge Gaps: Identifies unanswered questions and missing decision context.
• Knowledge Graph: Visualizes relationships across engineering knowledge.

How I built it

I built the prototype using Triggerware for custom GitHub data table generation, the GitHub API for engineering data ingestion, Next.js and TypeScript for the application, Supabase + pgvector for storage and retrieval, Cognee for connected knowledge, OpenAI for AI reasoning and embeddings, and Cytoscape.js for knowledge graph.

Challenges I ran into

• Turning fragmented GitHub discussions into meaningful engineering context.
• Generating trustworthy AI answers with verifiable citations.
• Connecting historical decisions to potential downstream risks and business ROI.
• Integrating multiple AI tools and data pipelines within a tight hackathon deadline.

Accomplishments that I'm proud of

• Built an end-to-end AI prototype independently using real LangChain GitHub data.
• Developed five connected features, going beyond traditional RAG-based search.
• Demonstrated how historical knowledge can support engineering decisions and human approvals.

What I learned

I learned that retrieving information is only the beginning. The real value comes from understanding decision context, evaluating risks, and connecting AI insights to measurable business outcomes. I also gained hands-on experience integrating multiple AI technologies under time constraints.

What's next for Company Memory Guardian

• Observability: Monitor AI traces, failures, retrieval performance, and system reliability.
• Latency: Track response times, including p50/p95 latency, and optimize performance.
• Tokenomics: Measure token usage and cost per request; optimize through caching, prompt efficiency, and model routing.
• AI Evals: Create golden data set and auto eval trigger setup.
• Enterprise Integrations: Expand continuous synchronization and connect decision intelligence to real GitHub PR approval workflows.

My vision: Turn company knowledge into better decisions, lower engineering risk, and measurable ROI.

Built With

• cognee
• cytoscape.js
• github
• llm
• next.js
• node.js
• python
• react
• supabase
• tailwind
• triggerware
• typescript