Software Engineer · U of T '28

DerrickRatnaharan

I'm an Applied Math & Statistics student at the University of Toronto. I've written a Llama inference runtime in C++, a static-analysis engine in Rust, and shipped a campus app to 300+ students.

View my work Resume

Open to 2027 SWE & ML internships

Derrick Ratnaharan smiling on a hike above a lake

300+

students using Campify

platforms shipped — iOS, Android, web

70+

AI patches evaluated at Mercor

3×

hackathon winner

Selected work

Selected projects

An inference engine, a code-analysis tool and a consumer app. Each one has a write-up with the architecture, the numbers and what I'd change.

CAUSAL MASK · KV CACHED

01LLM inference runtime

Kestrel

Llama-family inference in C++20 with CPU and Metal backends. No PyTorch, no llama.cpp. I wrote the forward pass myself and checked every layer against Hugging Face.

max logit error vs. HF: 4.1e-5; max logit error vs. HF
CPU decode, M2 Pro: 53 tok/s; CPU decode, M2 Pro

C++20
Metal
CMake
Safetensors

Case study

IMPACT ← CHANGED SYMBOL

02Static-analysis engine

RepoGraph

A Rust engine that uses Tree-sitter to map the calls and imports in a Python or TypeScript repo, then tells you which tests a change can break, without guessing.

cold index, 18K-LOC Flask: 0.36s; cold index, 18K-LOC Flask
graph edges resolved: 14,954; graph edges resolved

Rust
Tree-sitter
Graph algorithms

Case study

iOS · ANDROID · WEB

03Campus social app

Campify

A campus social app on web, iOS and Android. Students import their timetable from a screenshot, see when friends are free, and find people in their classes.

students onboarded: 300+; students onboarded
profile completion: 54→71%; profile completion

Flutter
React
Firebase
Node.js

Case study

Experience

Where I've worked

Oct 2026 — Present
Technical Lead · Google Developer Groups on Campus
- Build the technical content for hands-on AI/ML and software engineering workshops: live coding demos and project-based materials.
- Plan sessions on app development, debugging and deployment with the executive team.
Mississauga, ON
May — Sep 2026
Software Engineering Evaluator (Contract) · Mercor Intelligence & Handshake AI
- Evaluated 70+ software engineering tasks across Python, JavaScript, TypeScript and backend codebases. Reproduced failures and checked AI-generated patches against tests, edge cases and specs.
- Set up and debugged unfamiliar repos with Git, Docker and Linux to tell implementation bugs apart from environment, dependency and config failures.
Remote
Apr 2023 — Jun 2024
Software Team Lead · FIRST Robotics — Markham Robotics
- Led a 6-person team building a web platform for communication, strategy planning and match tracking. It cut competition prep by 2+ hours per event.
- Wrote Python scouting and analytics tools over 50+ match records, cutting manual data processing by 70%.
Markham, ON

Education

University of Toronto Mississauga

B.Sc. Applied Mathematics & Applied Statistics · Minor in Computer Science

Data Structures · Software Design · Object-Oriented Programming · Systems Programming · Probability & Statistics · Multivariable Calculus

Sep 2024 — Jun 2028

About

About me

I'm an Applied Math & Statistics student at the University of Toronto, and I spend most of my free time building software. What I enjoy most is going one level deeper than I strictly need to.

That instinct runs through the work above. With Kestrel I wrote an LLM inference engine from scratch to see exactly what happens between a prompt and a token. With RepoGraph I built a tool that answers “what does this change break?” from the code itself. And with Campify I shipped a product to 300+ real users and dealt with everything that comes with that.

I'm looking for a team where correctness and speed meet real products: software that's fast and right, and things people actually use.

Currently