Hello!

Forward-deployed engineer leading a team of software and ML engineers building AI-powered review of legal applications for a public sector org (OCR, agentic legal checking, synthetic data generation)

Led the design and implementation of core payroll features in Node.js (NestJS), PostgreSQL, and React—including attendance integration, formula-driven pay fields, and advanced role-based permissioning—powering £100M+ in annual payroll

Mentored junior engineers on writing efficient SQL queries, optimizing caching strategies, and handling production incidents—improving code quality and overall team capacity

Led backend engineering for an analytics platform in Python (Django). Grew ARR from 0 to $300k+.

Developed 10+ integrations for new data sources like QuickBooks, NetSuite, and HubSpot.

Built collaborative commenting feature in React (Next.js) and Python (Django).

Reduced backend response time by half by writing direct SQL queries.

Developed rental asset management, payments, and contracts for ~10,000 fleets using React and Go.

Resolved open production issues during on call. Saved ~$150,000 in potential missing payments.

Led agreement contract creation feature development with PDF generation and dynamic contract templates.

Developed a Node.js backend for employee recruiting, management, and attendance, serving 1M users.

Led candidate search and shortlist feature development in React Native.

Built barcode payment using Next.js and Node.js, significantly reducing errors from manual code entry. Increasing convenience-store-based transaction speed by 100% and success by 10%

Implemented BDD testing scenarios in Cucumber and Cypress along with a Slack reporter. Eliminated manual endpoint testing calls. Accelerating the development cycle by 20%

Built staging mode for bank connectors with Node.js, increasing confidence in testing and development

Integrated GitPrime into an internal platform for historical git data related insights

For my final-year research, I built an autograder that generates test cases by analyzing the semantic difference between reference and student implementations—achieving better coverage than random test generation. It has since been cited in multiple papers on automated grading.

Semantic-Based Autograding