Shiva – Python, React, Amazon S3

A central project-management system for an engineering consultancy, combining a React/TypeScript web application with Three.js for in-browser 3D model viewing, a C# backend on PostgreSQL running on Azure, and custom Autodesk Revit plugins that connect the platform with BIM authoring tools. The platform supports project management and BIM workflows across web and desktop environments.

• Build full-stack features end to end — from React/TypeScript UI through C# service endpoints to PostgreSQL schema changes — rather than working on one side of a front-end/back-end handoff;
• Develop Autodesk Revit API plugins in C#, extending the firm's BIM authoring environment and connecting it to the central platform;
• Own the Azure DevOps CI/CD pipelines: build, test, and release automation for both the web application and desktop plugins.

Builder Central is a project-management platform for construction teams, used to plan and track project work. The platform combines a React front end with Spring Boot and FastAPI backend services, supported by PostgreSQL and MongoDB for structured and document-based project data.

• Spearheaded the platform from the first commit — set the architecture, chose the stack, and drove the build across nearly five years of continuous development;
• Designed the backend as complementary Spring Boot and FastAPI services, allowing each to handle the workloads they suit rather than forcing everything through one runtime;
• Built the persistence layer across PostgreSQL and MongoDB, pairing a relational store with a document store to cover structured project data and less rigidly shaped records;
• Developed the React front end that construction teams use daily to plan and track project work;
• Maintained and evolved the system in production across nearly five years of changing requirements, rather than handing it off after initial delivery.

EK24 is an internal Autodesk Revit-based tool that automates kitchen design production for a construction business, generating construction drawings, renders, and a purchase-ready line-item list for manufacturer orders and site delivery. It replaced a manual, per-project drafting and ordering process with a repeatable automated pipeline, supported by a React interface and AWS/S3 for generated artifacts.

• Built EK24 end to end as its developer: the Revit integration itself, the React interface on top of it, and the AWS/S3 storage layer holding generated drawings and renders;
• Automated production of kitchen construction drawings and renders that had previously been drafted by hand for every project;
• Exported a purchase-ready take-off directly from the model — every line item needed to order from the manufacturer and ship to site — so procurement worked from the design itself instead of a separately re-counted list;
• Cut roughly two weeks from project timelines and about 4% from total project cost.

Automatic Quantity Take-Off (AQT) is a C# program that processes Autodesk Revit projects through a decision-rule engine to calculate quantities and generates formatted Excel workbooks with structured sheets, styling, and live formulas. Built from zero to one, it automates a previously manual quantity take-off process and provides a consistent, reviewable output for project teams.

• Built the take-off engine in C#: it reads an Autodesk Revit project and evaluates the model against decision rules to derive quantities, instead of relying on manual drawing interpretation;
• Encoded the estimating logic as explicit decision rules, making calculations consistent across projects and reviewable as code rather than buried in a hand-maintained spreadsheet;
• Generated finished Excel workbooks by constructing sheets, applying formatting, and writing formulas, so the result is a working document rather than a raw data dump;
• Took the project from an empty repository to production use and owned every architectural decision along the way;
• Automated quantity take-off that had previously been counted and reconciled by hand, cutting manual errors by 25%.

A developer-relations role at a developer-tooling startup, focused on internal engineering tooling and backend platform architecture. The main product was a Python service that automated multi-platform content distribution, using PostgreSQL for relational data and MongoDB for document storage.

• Designed and built a Python backend — PostgreSQL alongside MongoDB — that automated publishing articles across multiple platforms, removing a recurring manual step from the content pipeline. Owned it end to end, from data model through to per-platform publishing integrations;
• Made the technical case for moving the backend APIs to a microservices architecture and helped drive the transition, materially improving scalability and reliability;
• Ran UI/UX improvements based on direct user feedback, turning recurring friction points into concrete interface changes;
• Improved logistics efficiency by decreasing inventory administration time by 20%.