Back to projects

ESGTree Financed Emissions

An enterprise platform for calculating and reporting financed emissions for banks, portfolio managers and their SME portfolio companies.

Enterprise analytics dashboard for financed emissions reporting.
Invite SMEs and portfolio companies via secure invitation codes.
Activity log feature.
SME and portfolio company view.
SME and portfolio company wizard for guided data entry with AI parsing of documents.
Emissions estimation
File and bull uploads, powered by AI parsing and human-in-the-loop verification.
Summary of completed data entered by SME users, showing calculated emissions and data quality scores.

Overview

ESGTree is a multi-tenant SaaS platform that helps financial firms onboard SME portfolio companies, collect emissions data, and generate PCAF-aligned financed emissions reports. It serves two distinct personas through separate interfaces: a data-dense firm dashboard for analysts and a guided step-by-step wizard for SME users entering their own emissions data.

I led full AI-driven development of a full-stack MVP from prototype to production in under 4 weeks, guiding a team of 2 engineers and owning architecture and delivery. I made deliberate tradeoffs to prioritize speed and validation. The product won a competitive vendor demo with Desjardins, securing a paid pilot with onboarding in progress.

After validating product-market fit, I led a backend refactor to NestJS to introduce modular architecture, enforce clear boundaries, and improve reliability and data integrity. This included migrating to Drizzle ORM, adding session-based dual authentication (Replit OIDC and email/password), establishing a database-driven emissions activity model, and building a 400+ test suite covering unit, integration, and E2E flows — transitioning the system from a rapid prototype to a production-grade platform.

What I built

AI-Driven 0→1 MVP Delivery

Led full AI-driven development of a full-stack MVP from prototype to production in under 4 weeks, making deliberate tradeoffs to prioritize speed and validation. Won a competitive vendor demo with Desjardins, securing a paid pilot.

PCAF-Aligned Emissions Engine

Built a calculation engine for Scope 1, 2, and 3 financed emissions using the PCAF standard: Attribution Factor, Gross Emissions, and Data Quality Scoring. Supports spend-based proxy estimation for all 5 activity types when direct data is unavailable.

AI-Powered Document Parsing

SMEs can upload CSV, Excel, or PDF files to auto-populate emissions forms. A multi-stage pipeline extracts data via header matching or OpenAI, then applies post-AI validation with human-in-the-loop verification before acceptance.

Dual-Interface Product Design

Designed two tailored interfaces from a single backend: a data-dense firm dashboard for portfolio analysts and a guided step-by-step wizard for SME data entry — each optimized for its persona's workflow and context.

Modular Backend Architecture

Led a backend refactor to NestJS to introduce modular architecture, enforce clear tenant boundaries, and improve reliability and data integrity. Established a database-driven activity model so emissions categories can be updated without code changes.

Multi-Country & Bilingual Support

Supports 13 countries and bilingual (French/English) data entry. Emissions factors and sector data are seeded from Excel workbooks via a hash-gated system — changes to source data propagate automatically on deploy without code changes.

End-to-End System Ownership

Designed the multi-tenant data model, session-based dual auth (Replit OIDC + email/password), invitation flows, bulk import pipeline, and a 400+ test suite across 15 files covering unit, integration, and E2E scenarios.

Results

  • Led full AI-driven development of a full-stack MVP from prototype to production in under 4 weeks.
  • Won a competitive vendor demo with Desjardins, securing a paid pilot with onboarding in progress.
  • Transitioned the system from a rapid prototype to a production-grade platform by leading a NestJS backend refactor.
  • 400+ tests across 15 test files covering unit, integration, and E2E flows.
  • 5 emissions activity types across Scope 1, 2, and 3 with PCAF-aligned attribution calculations.
  • 13 countries supported via a hash-gated, Excel-driven seed system — zero code changes required to update emissions factors.