About Me.

Open-Source Contributor & Developer, SIG Apps & SIG Cluster Lifecycle

Feb 2026 - Present · 2 mos

Kubernetes (Official) • Remote

• SIG Apps (LWS) & SIG Cluster Lifecycle (kubespray) Team. Most recently contributed to LeaderWorkerSet (LWS), a Kubernetes SIG Apps–associated open-source project for managing groups of pods as a single unit of replication, designed for AI/ML inference workloads such as multi-host LLM serving. Worked on controller-level fixes in the Go codebase.

Open Source Research Contributor | llm-d (vLLM ecosystem)

May 2026 - Present · 2 mos

vLLM • Remote

• Open-source development for llm-d, a Kubernetes-native LLM serving framework, focusing on sustainable and carbon-aware AI infrastructure. Designed and engineered an Energy-Aware Endpoint Picker Plugin (EPP) to optimize token-level routing across heterogeneous hardware accelerators (e.g., H100s, L4s).
• Architected a Multi-Objective Routing Engine: Developed a high-performance Kubernetes Gateway API Inference Extension (GIE) sidecar in Go (v1.25), introducing a strict epsilon-constraint routing pipeline that optimizes for energy and carbon efficiency while strictly enforcing TTFT and TPOT latency SLOs.
• Phase-Aware Disaggregated Serving: Designed dynamic scoring models that differentiate between compute-bound prefill phases and memory-bound decode phases, actively routing requests to the most thermodynamically efficient nodes while calculating network KV-cache transfer penalties.
• Carbon-Aware Controller: Implemented real-time dynamic weight modulation based on grid carbon intensity (SCI standards) and cluster thermal budgets, enabling micro-spatial workload shifting during carbon spikes.
• High-Performance Go Engineering: Built a highly concurrent, thread-safe asynchronous eBPF telemetry ingestion plane using DCGM and RAPL metrics and Welford’s Algorithm for ms-precision signal filtering. Containerized the microservice into an ultra-minimal 8.6 MB distroless image with a negligible 101 microseconds routing overhead.
• Kubernetes CRD Reconciliation: Designed a native client-go Informer architecture to actively reconcile InferenceObjective CRDs, allowing cluster admins to dynamically pivot between autonomous “Carbon-Critical” and “Load-Shedding” control states.
• Quantifiable Impact: Achieved an average 17.4% reduction in LLM inference energy consumption (up to 32.3% for decode-heavy workloads) compared to standard round-robin scheduling, driving significant reductions in Scope 2 and Scope 3 data center emissions.

Machine Learning Research Assistant

Nov 2025 - Present · 5 mos

Deel • Remote

• Built a scalable legal-data RAG pipeline using Python and Selenium to automate the ingestion of 10,000+ CanLII case law documents into a Pinecone vector database, implementing structure-aware semantic chunking that reduced manual data curation time by 15 hours/week and LLM hallucination rates by 30% over baseline GPT-4 implementation.
• Developed a Random Forest classifier on 1260+ annotated employment-law cases to predict worker classification with 95% accuracy, leveraging Gini feature importance to provide legal interpretability for Sagaz test factors.
• Containerized RAG services and FastAPI using multi-stage Docker builds for deployment on Azure Kubernetes Cluster, ensuring 99.5% uptime through automated GitHub Actions CI/CD and liveness/readiness probes.

CAN Lead & Embedded System Engineer

Aug 2025 - Present · 8 mos

Queen’s AutoDrive Team • Kingston, Ontario, Canada

• Member of the Queen’s AutoDrive Embedded Systems Subteam. Designing an SAE Level-4 autonomous vehicle as part of the SAE & GM AutoDrive Challenge II competition by 2026.
• Architected and implemented a multi-threaded ROS2 C++ node integrating CAN messaging library with DBC parsing on an on-vehicle computing cluster (Raspberry Pi 5 + Central Compute) for an SAE Level 4 Autonomous Vehicle (Chevrolet Bolt), enabling real-time vehicle control commands (steering, braking, propulsion) of 100+ vehicle control signals with 10-50ms cyclic transmission rates across 4 isolated CAN networks (High-Speed, Chassis Expansion, Low-Speed, and Scoring CAN).
• Implemented ISO-TP diagnostic messaging system with XML-driven configuration, reducing diagnostic request setup time by 85% through automated signal encoding and protection value calculation.
• Developed lock-free MPMC queue-based signal routing system supporting 200+ concurrent CAN signals with zero-copy data transfer between ROS topics and vehicle networks, achieving sub-2 ms end-to-end latency.
• Built state machine framework with rolling counter protection and CRC validation for safety-critical messages (0x2CB, 0x315, 0x337), ensuring ISO 26262 ASIL-D compliance.

Sci '26 Vice-President, Equity Officer & ECE Discipline Club Treasurer

Sep 2024 - Present · 1 yr 7 mos

Engineering Society of Queen's University • Kingston, Ontario, Canada

• [Sci '26 VP]: Co-lead and manage a team of 12 executives for the Class of 2026, overseeing event planning for 800+ students while guiding strategic policy initiatives, advocating for student interests in Faculty Board meetings.
• Represent Sci ’26/Class of 2026 as a voting member of the Engineering Society (EngSoc) Council to ensure Sci ’26 voices are heard in key governance decisions affecting the 3,000+ engineering student body.
• [Equity Officer]: Embedded EDII-AS (Equity, Diversity, Inclusion, Indigeneity, Accessibility, Sustainability) principles into governance frameworks, policy development, and student leadership initiatives to ensure equitable processes across elections, awards, and council operations.
• [ECE Treasurer]: Served as ECE Discipline Treasurer managing a $19,000 operating budget, yielding a 30% increase in utilization efficiency. Managed all financial reporting, banking, reimbursement operations, and attained a balanced budget and a $9,000 net surplus by year-end.

Co-Founder, COO, HR Lead & Financial Lead

Dec 2023 - Present · 2 yrs 4 mos

Queen's High-Performance Computing • Kingston, Ontario, Canada

• Co-founded the first-ratified HPC club at Queen's and established its training and internal competition structural framework, helping students prepare for international Student Cluster Competitions (IndySCC/SCC).
• Delivered tutorials and workshops on HPC fundamentals, including cluster architectures (Frontenac), Linux system administration, parallelization (MPI/OpenMP), and benchmarking (HPL, HPCG, LAMMPS, Phasta).
• Implemented a 'Sprint' learning model with 4-week project cycles focused on distributed systems, Linux performance monitoring, and application tuning, creating lasting documentation for 40+ members.
• [HR Lead]: Led end-to-end recruitment cycle for the 2026-27 executive team using Breezy HR, conducting structured interviews and preparing relevant documentation for 20+ shortlisted candidates and ensuring EDII-AS policy compliance.
• [Financial/Sponsorships Lead]: Managed a $10,000+ operating budget preparing 40+ members to compete at SC25/SC26 and secured $10,000+ in funding from partners like Dell and Queen’s Centre for Advanced Computing. Automated purchasing analysis and optimized supply logistics using Excel-based workflow analysis, reducing procurement cycle time by 20% and 15% in overall expenses.
• Orchestrated the assembly and configuration of an 8-node Raspberry Pi cluster, configuring DHCP/TFTP for network boot and NFS for shared scratch disk management.
• Designed and developed 'travelExpenseTracker', an Android finance app with SQLite-backed authentication and real-time spending analytics using MPAndroidChart for competition teams.

FYIC Webmaster

Mar 2025 - Dec 2025 · 10 mos

Engineering Student Societies' Council of Ontario • Kingston, Ontario, Canada

• Spearheaded the end-to-end development, maintenance, optimization, and deployment of the official First Year Integration Conference (FYIC) website for over 18+ universities.
• Built a responsive, server-side rendered application using Next.js 15 and Tailwind CSS, achieving sub-second page load speeds and a 95+ Google Lighthouse performance score.
• Architected a robust GitHub Actions workflow to automate type-checking, linting, and zero-downtime production deployments to Vercel, reducing release cycles by 60%.
• Implemented automated vulnerability scanning within the CI pipeline to detect and patch critical CVEs in React Server Components.
• Integrated UserWay’s AI-powered accessibility widget to ensure WCAG 2.1 AA compliance.

Full Stack Developer

Sep 2024 - Apr 2025 · 8 mos

Queen's University Web Development - QWeb • Kingston, Ontario, Canada

• Designed and deployed a scalable full-stack e-learning platform using React.js, Node.js, Express.js, and MongoDB — serving 100+ monthly active users; managed state with Context API and leveraged Mongoose schemas for data modeling.
• Improved front-end performance through React.js optimizations like lazy loading and memoization, reducing page render times by 15% for critical user workflows.
• Built RESTful APIs with Node.js, Express.js, and MongoDB to handle CRUD operations, JWT-based authentication, and data aggregation for client projects, improving database query efficiency through efficient indexing and pipeline optimizations.
• Collaborated with 4 project designers and developers in Agile teams to build client websites, resolving 95% of identified bugs pre-deployment via automated testing (Jest); implemented Jest unit and integration tests for front-end components and back-end endpoints (via Supertest).
• Built responsive, WCAG-compliant UIs by translating Figma mocks into responsive, semantically structured React components with ARIA attributes and CSS modules; delivered full keyboard navigation and color contrast compliance.
• Streamlined development workflows using Git and managed CI/CD pipelines via Bitbucket, enabling seamless collaboration across 2 cross-functional teams (UI/UX designer & developer teams).
• Containerized the application using Docker and deployed it to AWS Fargate for scalable, production-grade hosting, reducing hosting costs by 30% through efficient container orchestration and RDS configurations.

Android App Development Freelancer

May 2024 - Aug 2024 · 4 mos

Freelancer.com • Remote

• Developed, designed, and delivered a feature-rich prototype e-commerce Android application (ChillnCharm) as a proof-of-concept for over 50 products.
• Leveraged MVVM architecture, the Repository pattern, and Clean Architecture to build a modular, scalable, and maintainable SDK codebase.
• Integrated remote and local data sources, using Retrofit to access the FAKE STORE API, retrieving over 50 product listings, and ROOM Database for local storage, achieving a 25% faster load time for cached content compared to API calls in benchmark testing.
• Created an intuitive and engaging UI that improved the shopping experience, including product search across 5+ categories, comprehensive cart management, and wish list persistence.
• Successfully completed and delivered the demo freelance project within deadlines, demonstrating strong project management skills and adaptability to clients' evolving requirements.

• Assisted in designing and implementing an assistive robotic feeding system integrating a Kinova Gen3 6-DOF collaborative arm onto a Clearpath Jackal UGV to autonomously transfer spoon-held food to physically disabled users.
• Developed a computer vision pipeline using Python OpenCV and ArUco markers to estimate 3D poses of food items for real-time trajectory offsets.
• Co-designed a modular ROS-based state machine to coordinate autonomous workflows with manual joystick overrides.
• Learned to implement Cartesian impedance controller integrating joint torque sensing for force-limited manipulation.

Project Financial Management Intern

May 2023 - Aug 2023 · 4 mos

Arista Networks • Hong Kong SAR

• Conducted financial modeling & viability analysis using Excel to support data-driven decision-making for $2 million worth of client pilot and internal automation portfolios.
• Designed real-time tracking dashboards in Power BI and flagged $150K+ in potential overruns on 6 active projects in SAP S/4HANA.
• Proposed and prototyped a Power Automate + SharePoint Dataverse workflow to streamline data collection, improving turnaround time by 20%.

Design Engineer

Jan 2023 - May 2023 · 5 mos

Queen's Relectric Car Team • Kingston, Ontario, Canada

• Designed a liquid-cooled battery enclosure for a 16 LiFePO4 batteries battery management system (BMS) in a Jeep TJ conversion.
• Modeled a 360mm × 637.5mm × 250mm aluminum 2024 alloy enclosure with SolidWorks, integrating copper tubing (10mm diameter) for heat dissipation.
• Conducting thermodynamic simulations and automating coolant pump activation via relay control to reduce thermal spikes by 40% compared to passive cooling.

High School Diploma

Sep 2018 - Jun 2022

Pierre Elliott Trudeau High School

• Top 6 Admission Average: 93% (English, Advanced Functions, Calculus and Vectors, Physics, Chemistry, and Mandarin)

Languages, Frameworks, and Libraries

HTML5

CSS

JavaScript (JS)

TypeScript 5.8

Tailwind CSS

Bootstrap

Material Tailwind

React.js & React Native

Next.js

Expo

Node.js

Express.js

Python

C

C++ & Embedded C++

Java

Kotlin

Go

Perl

R

SQL

MongoDB

Mongoose

MySQL

PostgreSQL

Django

Flask

FastAPI

Spring Boot

Google Gemini

LangChain

LangGraph

Pinecone

ChromaDB

scikit-learn

PyTorch

TensorFlow & Lite

Hugging Face

LLaMA & CTransformers

pandas

NumPy

Matplotlib

Seaborn

OpenCV

CVZone

Assembly language

ROS2

VHDL

SystemVerilog/Verilog

LaTeX

Bash/Shell

Framer Motion

GSAP

Three.js

Canvas Confetti

Lucide & Heroicons

React Hook Form

Vite

GraphQL

Webhooks

Context API

Redux

Selenium

BeautifulSoup

Jest

Supertest

OpenMPI

OpenMP

Room Database

Retrofit

Android Jetpack Compose

Next.js API Routes

Nodemailer (SMTP)

Sharp

PostCSS

Autoprefixer

ESLint

Developer Tools & Workflows

Kubernetes (SIG-Apps)

LeaderWorkerSet (LWS)

Kubespray & SIG Cluster Lifecycle

Docker

Helm

OpenShift Kubernetes

GitHub Actions CI/CD

GitLab CI

Bitbucket

Jira (Scrum)

Confluence

Jenkins

Ansible

Terraform

VMware

VS Code

Vim

JetBrains Suite

Android Studio

SolidWorks

Figma

Fritzing

KiCad

LTspice

Intel Quartus Prime, Quartus II, ModelSim

Grafana

Postman

MySQL Workbench

Slurm & Bash

Zephyr RTOS

CAN / ISO-TP

ISO 26262 (ASIL-D)

Google Cloud Build (GCB)

KIND (Kubernetes in Docker)

Dependabot

Vagrant

Google Lighthouse

WCAG 2.1 AA

Agile

Scrum

SDLC

Office 365

Google Colab

Jupyter

Databases and Enterprise Cloud Architectures

SAP S/4HANA Cloud

ORACLE Database

Microsoft Power BI

AWS (EC2, S3, Lambda, SQS, SNS)

AWS CloudWatch & Step Function

Azure & Azure Kubernetes Service (AKS)

Google Cloud Platform (GCP)

Vercel

Cloudflare

Supabase (PostgreSQL with RLS)

Firebase

Render

MySQL Database

PostgreSQL / SQLite

Redis

Prisma

Stripe

Excel (Dynamic Arrays & Stock Data)

SharePoint

Dataverse

Power Automate

Hardware, Systems & Networking

Raspberry Pi 5

PREEMPT_RT Linux

ValueCAN 4 & DBC Parsing

XML-driven config

MPMC zero-copy ROS–CAN transfer

ECU systems & CAN bus

SPI / I2C / UART

nRF52840 / ESP32 / Arduino / RF

Cyclone V FPGA

RISC-V ISA

Thread mesh networking

Coral TPU & EfficientDet

CUDA & MPI

DVFS & RAPL

OpenConfig

etcd & CoreDNS

Linux cgroup v1 deprecation

pod DNS & DPANIC crashes

Data, ML & Software Engineering Concepts

structure-aware semantic chunking

Random Forest & GridSearchCV

Gini feature importance

legal interpretability & LLM hallucinations

LLM-powered resume automation SaaS

memory-efficient RAG chatbot

quantized LLaMA models (.gguf)

context windows & token limits

salient object segmentation (U²-Net)

two-level nested U²-Net architecture

OpenCV morphological mask refinement

low-power autonomous object detection

MVVM Clean Architecture & Repository pattern

RESTful APIs & FAKE STORE REST API

API uptime & rate limiting

CI/CD & modular codebase

Redis rate limiting & caching layer

Caffeine & resilience4j

multi-stage Docker builds & containerizing

data preprocessing

offline functionality

CanLII (Legal Data Retrieval)

pdfplumber

Open-Source Contributions

vLLM (llm-d ecosystem)

Energy-Aware Endpoint Picker Plugin (EPP)

Multi-Objective Routing Engine

eBPF Telemetry & DCGM / RAPL metrics

Kubernetes 1.35 cgroup v1 deprecation

SIG Apps upstream maintainers

SIG Cluster Lifecycle (kubespray)

LeaderWorkerSet (LWS) controller fixes

Go controller-runtime patches

CNCF Open Source Workflow

Kubernetes Enhancement Proposals (KEPs)

Unit / Integration / E2E Testing

Ginkgo & Gomega test frameworks

Code Review & PR collaboration