Waterloo CS · automation · quality

I ship dependable software and strong automation — with machine learning when it actually moves the needle.

Computer Science (2B), University of Waterloo. I care about automation, testable systems, and full-stack delivery — and I use ML when it clearly improves the outcome.

Book a meeting Email me LinkedIn GitHub Resume (PDF)

Work

Experience

Automation Developer

Zoocasa · Toronto, Canada

Jan 2026 – Apr 2026

  • Improved release predictability by building end-to-end test automation (Playwright, Postman) across JS and GoLang services, cutting manual QA effort 60% and shortening feedback loops for frontend and backend product teams.
  • Shipped an AI-assisted compliance workflow: natural-language answers over policy docs, structured JSON configs generated from source files, and real-time updates on document changes, reducing manual review and clarifying impact for GTM stakeholders.
  • Ran structured post-incident reviews on production defects across codebases; prioritized fixes that cut repeat incidents in high-risk areas and lifted reliability 47% (measurable defect trend).
  • Partnered in sprint planning, sizing, and feature prioritization with PM, Design, and Engineering, translating goals into shippable increments and keeping delivery dates honest.

QA Specialist

DocTalk · Toronto, Canada

May 2025 – Aug 2025

  • Defined and documented 1500+ functional requirements and test cases (edge cases, failure modes), single source of truth between Product and Engineering for scope, acceptance, and release readiness.
  • Owned B2B SaaS web work end-to-end: shipped clinician/admin-facing improvements, closed UX gaps from support feedback, and validated flows with clinical stakeholders before release.
  • Built automated regression coverage (1000+ scenarios, REST + UI, Playwright/Postman) so teams could ship with clearer risk picture and fewer production surprises.

Market Analyst Consultant

Happy Sneeze · USA, Remote

Apr 2024 – Jul 2024

  • Produced a competitive landscape and financial benchmarking for 15+ health-market players; ETL and analysis pipelines turned raw filings and metrics into exec-ready product and positioning implications.
  • Sized market segments and growth trends via reporting and scenario analysis; delivered concise memos that informed go-to-market and strategic bets at leadership level.
  • Recommended 10 partnership targets and built a structured outreach narrative (value props, talk tracks) aligned to growth goals, supporting BD conversations from first touch to follow-up.

Selected work

Projects

RefiMatch

Refinance scenarios ranked with AI

Apr 2026

Languages

  • Swift
  • Python

iOS & client

  • SwiftUI
  • iOS
  • Swift Package Manager

Backend & APIs

  • FastAPI
  • Pydantic
  • REST APIs

Data & ML

  • Pandas
  • NumPy
  • LangGraph
  • LangChain
  • LLM
  • RAG
  • pgvector
  • PostgreSQL

Platform & delivery

  • Docker
  • Kubernetes
  • Kustomize
  • GitHub Actions
  • CI/CD
  • Redis

Quality & observability

  • pytest
  • Ruff
  • observability
  • Langfuse
  • MCP
  • Promptfoo
  • DeepEval
  • Architected a monorepo delivering a SwiftUI iOS 17 client and a FastAPI backend that compares curated refinance offers using a Pandas/NumPy amortization and ranking engine, keeping numeric truth separate from model prose.
  • Implemented a LangGraph-based explanation flow with RAG retrieval over bundled knowledge and optional pgvector storage, plus Langfuse-compatible tracing hooks for production observability.
  • Exposed lender catalog and scoring operations through an MCP (Model Context Protocol) stdio server for reusable tool access from AI hosts and demos.
  • Shipped Kubernetes-first deployment artifacts (Postgres/pgvector, Redis, Job-based ingestion, API Deployment/Service) and GitHub Actions CI with Ruff, pytest, Promptfoo, and DeepEval hooks.
  • Built REST contracts with Pydantic v2 validation, JSON schema for compare payloads, and a Docker image bundling API, agent, and RAG ingest for repeatable environments.
  • Monorepo layout: separate scoring library, agent package, API service, RAG ingest, MCP server, K8s manifests, and eval harness—~80 automated pytest cases (scoring + agent graph + optional DeepEval), 6 Swift source files in the iOS app target.

OmenAI

Insider-trading energy on prediction markets — TD Best AI hack to detect financial Fraud

Mar 2026

Languages

  • Python (scikit-learn)

Frameworks

  • Next.js

Data & storage

  • PostgreSQL
  • Redis
  • Framed insider-risk as an analyst-facing product problem on prediction markets; scoped an MVP around unlabeled trade streams and tiered risk signals (rules + Isolation Forest) with narrative explanations for triage.
  • Prioritized 600K+ live trades into wallet-level features (velocity, sizing, win-rate drift) and segment-specific thresholds, balancing false positives vs. investigation load for demo users.
  • Shipped a live operations dashboard (streaming feeds, on-chain context, radar views, wallet drill-down) for TD Best AI / GenAI Genesis 2026 (Gold sponsor winner).

Weathwise

Money APIs with trust issues—in a good way

Dec 2025

Languages

  • Java
  • TypeScript

Frameworks

  • Spring Boot
  • Angular

Data & storage

  • PostgreSQL
  • MongoDB

Security

  • JWT

Testing & infrastructure

  • Postman
  • Docker
  • Defined security and data-isolation requirements for a personal-finance style workflow (JWT, per-user row scoping, schema versioning), translated into API contracts the web client could ship against.
  • Scoped OLTP vs. analytics paths (PostgreSQL vs. MongoDB) and failure-isolated batch slices so one bad account could not block aggregate jobs, with explicit tradeoffs for reliability vs. cost.
  • Delivered a containerized MVP (Docker Compose, health checks, reverse-proxied SPA) with documented APIs and automated tests on auth and aggregation, repeatable path from local to demo.

eXplicit

Compliance docs that answer back

Jan 2026

Languages

  • Python

APIs & integrations

  • Google Drive API
  • OpenAI

Data & automation

  • JSON
  • Workflow automation
  • Reduced manual policy review by shipping natural-language Q&A over real-estate feed rules with answers grounded in source docs, clear user value for ops and compliance reviewers.
  • Owned the document-to-config pipeline: ingestion from Google Drive, structured JSON artifacts per jurisdiction/profile, and removal of spreadsheet handoffs that caused versioning errors.
  • Implemented change notifications on Drive edits to regenerate configs, keeping downstream validators aligned with the latest policy without silent drift.

Squasher

Your repo's vibe check

Jan 2026

Languages

  • Python
  • TypeScript

Frameworks & platforms

  • FastAPI
  • LangGraph
  • VS Code API

Data & storage

  • ChromaDB
  • SQLite

DevOps & quality

  • Docker
  • pytest
  • Shipped a split client–server quality platform: FastAPI service exposing file-, repo-, and pre-commit-oriented analysis endpoints, paired with a TypeScript VS Code extension that surfaces risk dashboards and one-click re-runs from the editor.
  • Orchestrated analysis as a LangGraph state machine—Radon-backed complexity, git-blame velocity signals, static import graphs, and ChromaDB vector retrieval over curated anti-pattern corpora—fused into a 0–100 decay score with severity bands and remediation text.
  • Operationalized the backend for real teams: SQLite metric history, embedding lifecycle in Chroma, WebSocket progress fan-out for long scans, Dockerized deploys with auth and health probes, and pytest suites locking agent contracts and API regressions.

Saiborg

Talk, type, search, schedule—one assistant that remembers all

Nov 2025

Languages

  • Python
  • JavaScript
  • HTML
  • CSS

Frameworks & runtime

  • FastAPI
  • Electron
  • Uvicorn
  • Implemented a FastAPI-centered assistant shell with a browser SPA: Web Speech–driven voice capture, typed chat transcripts, and a search plane that prefers Google Programmable Search (with DDG-style fallback) then optionally routes hits through Claude or Gemini for grounded, link-preserving summaries.
  • Integrated Google Calendar end-to-end—OAuth desktop flow, token refresh, and CRUD helpers for upcoming events and quick-add flows—plus an Electron wrapper that supervises uvicorn, exposes a frameless desktop shell, and deep-links file pickers where needed.
  • Production-minded configuration: `.env`-driven API keys for search and LLM providers, optional Spotify OAuth for “now playing” cards, Vercel-oriented deploy notes, and documented cold-start paths for both web-only and Electron-first users.

Smart Mirror — AuraGlass

Mirror, mirror on the wall

Oct 2024

Languages & scripting

  • Bash
  • JavaScript

Hardware

  • Raspberry Pi 4

APIs & integrations

  • OpenAI

DevOps

  • GitLab CI
  • Led firmware-adjacent productization on Raspberry Pi 4: modular JavaScript services behind a Bash-orchestrated boot path, integrating Google Calendar, Spotify, and OpenAI-powered glanceable cards with GitLab CI for repeatable builds.
  • Owned module architecture and integration testing across seven mirror surfaces—latency-budgeted UI loops, API quota handling, and graceful degradation when third-party endpoints throttle or fail.
  • Ran the team through specification, integration spikes, and demo hardening; shipped a research-grade mirror stack benchmarked at ~90% scenario pass rate on acceptance scripts.

Running Alarm Clock — Alarmy

The snooze button's nemesis

Sept 2023

Languages

  • Java

Embedded

  • Arduino

Tooling

  • Git
  • Engineered a Java-hosted verification harness around an Arduino “runaway” alarm firmware: serial protocol tracing, timed actuator assertions, and Git-tracked fixtures so hardware regressions were reproducible on the bench.
  • Authored 20+ automated and semi-automated cases spanning wake windows, snooze edge cases, power-loss recovery, and motor-drive fault injection—tightening loop latency until the clock hit an 80% measured pass rate on the full matrix.
  • Published traceable QA artifacts—15+ documented cases with inputs, oracle behavior, and captured traces—for regression sign-off and performance baselines ahead of milestone demos.

Academics

Education

University of Waterloo

Bachelor of Computer Science - 2B

Waterloo, Canada · Sept 2024 – Present

Toolkit

Skills

Product & collaboration

  • Agile
  • JIRA
  • Figma
  • Excel
  • Tableau
  • Power BI
  • Requirements & acceptance criteria
  • User stories
  • SDLC
  • Cross-functional delivery
  • Financial modeling
  • Mortgage / refinance analytics
  • Decision-support systems

Languages & runtimes

  • Python
  • Java
  • C++
  • GoLang
  • JavaScript
  • TypeScript
  • Node.js
  • Swift

Mobile

  • SwiftUI
  • iOS
  • Swift Package Manager

Python & data

  • Pandas
  • NumPy
  • Matplotlib
  • SciPy
  • PyTorch
  • scikit-learn
  • ETL
  • Isolation Forest
  • LangChain

Web & APIs

  • Spring Boot
  • Angular
  • FastAPI
  • Uvicorn
  • Next.js
  • Astro
  • Vite
  • Tailwind CSS
  • Electron
  • REST APIs
  • JWT
  • OpenAPI
  • WebSockets
  • OAuth 2.0
  • Pydantic

Datastores

  • PostgreSQL
  • MongoDB
  • SQL
  • Supabase
  • Redis
  • SQLite
  • ChromaDB
  • pgvector

AI & integrations

  • OpenAI API
  • Google Drive API
  • Google Calendar API
  • Gemini API
  • Anthropic API
  • Web Speech API
  • Programmable Search API
  • LangGraph
  • LLM
  • RAG
  • Langfuse
  • MCP
  • Promptfoo
  • DeepEval

Platform & delivery

  • Git
  • GitHub
  • GitLab CI
  • Docker
  • Vercel
  • CI/CD
  • GitHub Actions
  • Kubernetes
  • Kustomize
  • Monorepo
  • observability
  • DevOps
  • Linux
  • Bash
  • Shell scripting
  • AWS
  • Azure
  • Maven
  • JUnit
  • Flyway
  • Karma
  • GTest

Quality & tools

  • Postman
  • Playwright
  • pytest
  • Ruff
  • VS Code API
  • Radon
  • Arduino
  • Raspberry Pi
  • DocketQA

Beyond the desk

Leadership & community

Events Director

UW Product Management Club · Waterloo, Canada

Sept 2025 – Dec 2025

  • Led end-to-end event planning and execution for UW PMC, from ideation and run-of-show through follow-up, balancing sponsor needs, venue constraints, and attendee experience.
  • Partnered with marketing on campaigns that lifted attendance and engagement; aligned messaging with club leadership goals.
  • Delegated setup/teardown and volunteer roles for small- and large-scale events; de-risked day-of logistics with contingency checklists.

Campus Ambassador

Velocity · Waterloo, Canada

Sept 2025 – Dec 2025

  • Managed event-day operations, including check-in, food stations, supply distribution, wayfinding, and attendee support.
  • Promoted campus events and programs through social media campaigns and word-of-mouth outreach.
  • Organized social activities to engage the Velocity community and strengthen campus connections.

Recognition

Certifications & awards

  • Completed AWS Beginners certification and Data Analysis Certification by LinkedIn, API Postman certification.
  • Won Gold Sponsor TD Best AI Hack to detect financial fraud at GenAI Genesis hackathon 2026.

COFFEE CHATS · RECRUITING · COLLABS

Let’s connect

For recruiting, mentorship, or collaboration — grab a slot on my scheduling page, or email me if you prefer to start there.

Book a call Email me