Kavya Kadi

Kavya Kadi · Software engineer · Toronto

I teach software to read financial documents

Software Engineer who takes complex ideas from first design to reliable production systems, working across backend engineering, distributed systems, data, and applied AI.

See the work kavyakadi@gmail.com

About Me

Portrait of Kavya Kadi

I’m a software engineer who turns complex, manual workflows into reliable systems.

I have professional experience at Manulife, ADP, and AMD, working across backend engineering, distributed systems, data platforms, and applied AI. I focus on building production software that improves reliability, simplifies operations, and delivers measurable results.

I’m especially interested in the engineering behind dependable AI products: clear APIs, resilient workflows, thoughtful validation, and the infrastructure that moves a system beyond the prototype stage.

The foundation: Computer Engineering at the University of Toronto, with minors in Artificial Intelligence and Engineering Business. Here, I also built a public-health analytics dashboard as a researcher under Dr. Yu Chen.

Professional Experience

  1. 2024 — now

    Manulife · Software Engineer, Investments AI

    Documents in, decisions out. I build production AI systems for investment operations, including multi-agent document workflows, validation APIs, event-driven backends, and cloud migration tooling. The goal is straightforward: reduce repetitive processing, improve reliability, and give analysts more time to review the work that requires judgment.

  2. 2023

    ADP · Software Engineering Intern, Lifion Payroll

    Payroll systems have to keep up with complex and frequently changing rules. At ADP, I built Node.js backend services and extended internal SQL tooling to automate tax-jurisdiction synchronization across the Lifion platform, replacing manual configuration and improving payroll data consistency.

  3. 2022 — 23

    AMD · Software Engineering Intern, Data Center APU (MI300)

    Firmware validation generates a large volume of data, and finding the right failure quickly matters. At AMD, I built Python debugging and regression automation and developed Airflow pipelines to process and surface results faster for engineering teams.

Selected Projects

LexiMatch AI 1st place · Manulife hackathon

LexiMatch uses Python, Flask, and NLP to compare both documents, surface clause-level changes, and turn a full legal re-read into a faster, more focused review.

Sheet music being processed by the optical music recognition pipeline

Optical Music Recognition

Sheet music is a document too. Built with Python and PyTorch, this computer-vision pipeline reads scanned scores and converts centuries-old musical notation into a machine-readable format.

Read the code
Voyager Mapper rendering an interactive city map

Voyager Mapper

A GIS application that renders city maps, surfaces points of interest, and computes practical routes. Built close to the metal with C++ and GTK, it turns geographic data into an interactive model of the city.

Read the code
Languages
Python · TypeScript · SQL · C++
Build with
FastAPI · Node.js · LangChain · PyTorch · Azure OpenAI · PostgreSQL · MongoDB · Airflow · Databricks
Ship on
Azure (AKS, API Management, Key Vault) · Docker · GitHub Actions · Jenkins

Currently reading: your message.

Interested in collaborating or discussing an opportunity? Let’s connect.

Write to me Find me on LinkedIn

Toronto, ON · github.com/KavyaKadi3