Nesta-related data platforms

Overview

Infomerx supplied embedded software engineering across several open-source programmes in the Nesta data-and-innovation ecosystem. Map-first UX—declarative layers, filters, clustering, and map–list coherence—was central to that work (especially on WAIfinder), not a side track.

WAIfinder

Desktop

Mobile

Nesta and Innovate UK collaboration: UK AI ecosystem explorer built on @svizzle/mapbox — Supercluster clustering with expansion-zoom on click; auto-zoom that disables on manual pan and re-enables to clear bbox area filter; OR / AND org-type filter toggle (supporting multi-category orgs); org pinning (hover = ephemeral preview, click = persistent detail panel); org–place–topic–region views and partner branding integration.

Live site · Source

Svizzle

Open-source Svelte data-visualisation toolkit published under nestauk and developed in parallel with programme use — consumed by WAIfinder, EURITO indicators, and Heat Pump Market Tracker. Contributions include: @svizzle/mapbox (Mapbox behind Svelte components — declarative map layers and map UI widgets, used heavily in WAIfinder); ScreenSensor (measures a rendered text sample via ResizeObserver to derive glyph width/height, then expresses layout breakpoints in character-count rather than pixels — so charts and panels adapt to the active font, not a fixed pixel assumption); A11yMenu (in-page accessibility panel: font family + scale, line height, letter/word spacing, contrast, brightness, hue rotation, grayscale, and CVD simulation for protanopia / deuteranopia / tritanopia — all persisted in localStorage); plus charts, histogram, choropleth, legend, UI primitives, and package documentation.

Documentation · Source

Heat Pump Market Tracker

Nesta ASF programme: geographic dashboard tracking UK heat pump installations. Features include map–chart cross-linking (bar hover highlights the map region and vice versa), filter and metric navigation, Svizzle's ScreenSensor + A11yMenu integrated with custom font families, pa11y / Lighthouse audits, and backend query-shape adapters plus a dev/staging → production data pipeline.

Source

Dapsboard

DAPS data exploration and query builder. Builder route: a custom runtime type system (xorOf, enumsOf, arrayOf, objectOf, optional, …) describes every ES aggregation's parameter schema; TypedField dispatches each type descriptor to the correct Svelte widget (BooleanRadios, RadioList, ArrayEditor, ObjectEditor, SimpleField) so forms generate entirely from the aggregation × field-type selection — no hand-authored form per aggregation. Explore route: XState parallel machine (History / Selecting / Results / Search); neededFields exists-query wiring; Fastify proxy to DAPS ElasticSearch with MongoDB hash-based response cache; coverage queries; dataset specs; and end-to-end authentication (Basic Auth via dap-tools.uk, CORS-gated unauthenticated access, protected datasets, HTTP 401 client handling).

Source

EURITO indicators

Application modernisation (Sapper to SvelteKit); the app is a thin configuration wrapper over time_region_value Svizzle routes — NUTS2 choropleth and time-series indicator views consumed wholesale as packages, demonstrating Svizzle's completeness as a reusable programme toolkit; guide routes and quality-tooling hooks.

Source

Annotation / infrastructure

Two-tier AWS system. Base layer: an interactive CloudFormation deploy script (validates R53 hosted zone, S3 cert, AMI; prompts config; polls stack events to CREATE_COMPLETE; confirms service readiness) that provisions four EC2 instances — API gateway, authentication, provision, and annotation coordinator — each with Route 53 DNS records derived from their public IPs and Certbot HTTPS. Scaling layer: the provision server exposes a REST API (/create, /teardown, /status) — POST /create { workers: N } generates a Terraform config with N annotation-node-* EC2 instances, runs terraform apply, reads public IPs from Terraform state, regenerates the nginx upstream block to load-balance across them, reloads nginx, and SSH-launches the DBpedia Spotlight Docker container on each new node. dap_dv_backends_utils provides the Terraform command wrappers and state reader imported by this service, as well as the NER annotation pipeline (ES scroll → batch → Spotlight → bulk update), SigV4-signed ES requests, S3↔ES sync, auth, and SES utilities — also consumed by Dapsboard.

dap_dv_backends source · dap_dv_backends_utils source

Challenge

Policy and research audiences need maps, indicators, and dashboards that stay legible under real metric density: coherent map–list behaviour, progressive disclosure instead of "chart soup," and reusable building blocks so each new tool does not reinvent visualization, auth, and deployment patterns.

Approach

Infomerx worked inside existing programme repos: ship vertical slices in the product repositories, land shared abstractions in the library where they belong, and keep operator-facing docs and scripts honest so pipelines stay deployable — not only demo-ready.

Outcomes

Outcomes are best read as capability and maintainability: reusable visualization and map primitives, production-shaped auth and explore flows, and pipelines that other researchers and engineers could extend without a ground-up rewrite. Any quantitative programme metrics stay with the programme owners and contracts — not reproduced here.