K·NI // NETRUNNER OS
TORONTO NODE// SYS.STATUS: ONLINE// CLASS OF 2026

KEVINNI

ROBOTICS ENGINEER · ML · QUANT

Final-year Engineering Science (Robotics) at the University of Toronto, operating at the seam where autonomous systems, machine learning, and quantitative finance collide.

ID // 0302
OPERATORKevin Ni
FACTIONUofT EngSci
DISCIPLINERobotics
SUBCLASSEng Business · AI
CLEARANCESkoll Scholar
LOCATIONToronto, ON
01

OPERATOR PROFILE

I'm a final-year Engineering Science student majoring in Robotics, with a double minor in Engineering Business and Artificial Intelligence. My foundation is hardware and control, but my trajectory points squarely at AI-driven quantitative finance — turning models into decisions that move capital.

My undergraduate thesis at the FORM Lab under Professor Roy Kwon fuses machine learning with convex optimization for statistical-arbitrage Markowitz portfolios — a pipeline that cleared survivorship bias on real CRSP/WRDS data and held its edge through live market regimes.

Before finance, I logged serious time in the field: autonomous drones and self-driving stacks on ROS2, embedded systems on bare-metal STM32, and a co-op designing electrical systems for 100+ interiors. Different domains, same instinct — drop into the unknown, learn fast, ship something that works.

Off the grid: certified deep-water scuba diver (dove the Galápagos with sharks and sea lions), 10+ years on the flute, plus weightlifting, cooking, and a growing shelf of Gundam model kits.

0.000
Peak Thesis Sharpe
$0
Skoll Award
Off-Duty // Background Processes
  • scuba_diver.sysGALÁPAGOS
  • flute.daemon10+ YRS
  • gundam_build.exeRUNNING
  • golf + badmintonIDLE
  • photography.capACTIVE
  • kitchen.sandboxFOODIE
02

SYSTEM LOG

2025 — 2026

FORM Lab — Roy Kwon

Undergraduate Thesis Researcher
+

ML-enhanced statistical-arbitrage on a Markowitz portfolio — figuring out where learning actually adds value in a quant pipeline.

  • Benchmarked 6 models across pair selection and parameter estimation; found ML pays off at estimation, not selection.
  • Built an ElasticNet + RandomForest → daily QP pipeline over 12,668 CRSP equities; hit Sharpe 0.661 and +49.4% return.
  • Killed survivorship bias via CRSP/WRDS (19.4% survival vs. ~93% for free data), exposing a 73 pp backtest inflation.
  • Surfaced regime dependence: +18.5% Sharpe vs. equal-weight in bull markets, drawdown capped at −14.1% in 2022.
PythonQP / Convex OptRandomForestElasticNetCRSP / WRDS
2024 — 2025

Smith + Andersen

Junior Electrical Designer
+

Dropped into building systems with zero prior background and made it my domain — coordinating architects, contractors, and budgets at consulting pace.

  • Designed electrical and lighting systems for 100+ interiors, from 500 sq ft retail to 60,000+ sq ft offices.
  • Produced 100+ AutoCAD / Revit drawing sets at a one-week average delivery turnaround.
  • Ran pricing and material-quote reviews to keep designs cost-accurate and on-budget.
AutoCADRevitCost ReviewCode Compliance
2024 — 2025

aUToronto — Self-Driving Team

Systems · Drivers
+

Sensor and driver work on the university's autonomous-vehicle stack, the perception layer's plumbing.

  • Automated camera-driver config in C++ / ROS2, cutting setup time 30% and steadying perception calibration.
  • Streamlined sensor integration with perception and controls for faster testing and higher uptime.
  • Annotated and validated 2,000+ training images for lane and object detection.
C++ROS2PerceptionSensor Fusion
2023

Kevares Autonomous Services

Jr. Mechatronics & Embedded Systems
+

First robotics internship — startup tempo, code on day one, real systems in the field.

  • Built a ROS + Streamlit dashboard wiring live telemetry, diagnostics, and control into one interface.
  • Wrote a Python stereo-vision backend (Open3D) for automated defect detection, cutting analysis time 40%.
  • Deployed and tested embedded systems on autonomous platforms, tightening latency and reliability.
ROSPythonOpen3DStreamlit
03

DEPLOYED BUILDS

BUILD_012025

Markowitz Portfolio Engine

Python · MATLAB · pandas · NumPy

Solved the mean-variance optimization problem for an n-asset portfolio to trace the full efficient frontier, with data pipelines computing 10-year returns, covariances, and volatilities.

Iterative QPoptimal weightsLong / Shortboth constraints
BUILD_022025

Audio Diffusion Transformer

Python · PyTorch · NumPy

A diffusion-transformer audio generator with test-time training, plus adaptive MLP blocks across 24 layers for real-time refinement and steadier generation.

+140%CLAP vs. baseline18.5 hrsaudio pipeline
BUILD_032025

Autonomous Tracking Drone

ROS2 · NVIDIA Jetson · YOLOv8 · MAVROS

A 7-node ROS2 system on Jetson with HSV / YOLOv8 vision and a 6-state flight FSM, including Archimedean-spiral re-acquisition on target loss via T265 visual-inertial odometry.

5–10 msdetection latency20 Hzcontrol loop
BUILD_042024

STM32 Robotic Companion

C · STM32 HAL · PWM · FSM

An interactive robotic system on STM32-DISCO blending a touchscreen UI, audio, sensors, LEDs, and servos, driven by an interrupt- and timer-based finite state machine.

Real-timeI/O syncHALbare-metal control
BUILD_052023

Equity Valuation Models

FactSet · Excel · DCF · CAPM

DCF and CAPM valuations for Caterpillar and Deere, with IRR / NPV analysis on a $120M project across tax and discount-rate scenarios, and DuPont risk comparisons.

$120Mproject modeledDuPontleverage analysis
BUILD_062024

This Site // v2

Three.js · WebGL · GLSL · Vanilla JS

A hand-built WebGL experience — infinite neon grid, a glowing cyberware core, bloom post-processing, and a HUD interface. No frameworks, just shaders and intent.

WebGL2+ bloom pass0dependencies
04

ARSENAL

Languages

PythonC / C++JavaSQLMATLAB

ML / AI

PyTorchscikit-learnNumPypandasDeep LearningDiffusion Models

Quant / Finance

Portfolio OptimizationConvex OptimizationRisk ModelingDCF / CAPMStat-ArbFactSet

Robotics / Systems

ROS2Embedded / STM32Control TheoryState EstimationComputer VisionNVIDIA Jetson
Establish Connection

LET'S BUILD

Open to quant, ML, and robotics roles starting 2026. If you're working on something at the intersection of intelligence and capital — let's talk.