kReative Labs
Back to kReative Labs

Live 2026

Knowledge graph

A resume is a list. This is the same career, ventures, projects, and writing rendered as a graph, with typed, reasoned edges instead of bullet points.

Explore the graph

The idea

A career is not actually a list

A resume forces everything into one sequence: role, then role, then role. That hides the more interesting structure underneath it. The multi-agent orchestration work at Alexa did not just precede the Noesis advisory role, it directly shaped the argument for treating AI agents as a managed workforce. unslop-text's two-pass design and Book writing agents' Noesis gate are not related because they shipped in the same year; they are related because they landed on the identical pattern (a deterministic check in front of a subjective one) independently.

The graph exists to make that structure visible instead of implied. Every node and edge in it traces back to something already public on this site: /profile.json, the /klabs/ project pages, and /blog.html. Nothing was invented for the graph. It is a different lens on the same facts, not a new set of them.

Architecture

Same facts, rendered two ways

Knowledge graph architecture The personal site's structured data (profile.json, the klabs project pages, and the blog) is manually distilled into a static data.js file of typed nodes and reasoned edges. A D3.js force simulation lays out that data as an SVG graph in the browser: colored dots for nodes, colored labeled lines for relationships. Hovering a node or edge highlights its connections and shows the reason for the link. Clicking a node with a URL opens the real page on the personal site in a new tab. profile.json klabs pages blog.html the site's own facts data.js typed nodes, reasoned edges D3 force layout runs entirely in the browser, no server, no database You hover for the reason click to open the real page
The only manual step is distilling site data into data.js. Everything after that, the physics, the coloring, the hover and click behavior, is D3 and plain SVG, no framework.

Edges carry one of four types: progression (chronological), informs (an experience shaped a later thing), built on (a project literally uses or extends another), and related (a looser thematic link). Each edge also carries a one-sentence reason, visible on hover, so a connection is never just an unexplained line.

Detail

At a glance

StackStatic HTML/CSS/JS, D3.js v7 for the force simulation and SVG rendering. No build step, no database, no API keys.
Data32 nodes across six categories (education, career, venture/advisory, kReative Labs project, writing, theme/skill) and 35 typed, reasoned edges, hand-distilled from the site's own structured data.
InteractionDrag to reposition, scroll to zoom, hover a node or edge to focus its connections and read the reason, click a node to open the real page on shridharpathak.me.
Source of truthEvery fact traces back to /profile.json, a /klabs/ page, or /blog.html. The graph is regenerated by hand when those change; it is a view, not a second database.
Repogithub.com/shripathmit/knowledge-graph
StageLive.