Codensity — Developer Sustainability Platform

Designing a deterministic, explainable sustainability calculation engine without AI or ML.

Ingesting batched SDK telemetry reliably, including retries and duplicate deliveries, without double-counting.

Implementing a multi-tenant architecture with organization, project, environment and SDK-key isolation.

Securing privileged admin operations on a platform that holds customer telemetry.

Built a three-step formula engine (compute units -> energy kWh -> CO₂ grams) using published coefficients, with the arithmetic exposed behind every figure.

Introduced pluggable ingestion sinks, so events are written directly to PostgreSQL by default or streamed through Kafka, deduplicated by a client-generated event ID.

Built a node-cron scheduler that recomputes hourly and daily rollups from the deduplicated raw table.

Added environment-scoped SDK keys with Redis-cached lookups and immediate invalidation on revocation.

Protected the super-admin area with TOTP MFA, Redis-backed sessions, re-authentication for sensitive actions, CSRF and origin guards, and audit logs.

Drop-in Node.js SDK (@codensity/node) with track() and record() APIs, batching, buffering and retries, plus Express middleware and a NestJS module, interceptor and decorator.

Sustainability dashboard showing CO₂ grams, energy kWh and compute units per feature.

Organizations, member invitations, and project/environment management with scoped SDK keys.

Transactional and weekly sustainability-report emails built with React Email and Resend.

Super-admin console with audit logs and MFA-protected sessions.

Product analytics with PostHog, and Stripe billing through Better Auth.

Published @codensity/node to npm with full TypeScript support, Express middleware and NestJS integration.

Launched a public beta at getcodensity.com with an end-to-end pipeline from SDK event to ingestion, aggregation and CO₂ calculation.

Selected as one of 6 finalists from 32 applications at the Innovatum Science Park startup competition (2026).

The complexity of translating raw compute metrics into meaningful sustainability signals for non-specialist audiences.

Designing SDK ergonomics that are zero-config for simple use cases but fully configurable for advanced integrations.

When a queue is the right tool and when scheduled batch work is simpler, which led to running aggregation on a cron scheduler.

Carrying a full rebrand (Carbonly to Codensity) across packages, Docker images, the npm SDK and live infrastructure.