Simarjot Singh
Software Engineer · Math and Computer Science at SFU
Projects
QuorumDB0 lost writes in 100 chaos tests
Redis-compatible key-value store in Go, replicated across three Google Cloud zones with Raft I wrote from scratch.
cluster-scheduler85% CPU use
Job scheduler in Java and gRPC that packs 10K jobs onto 100 nodes and never places a job twice, even across 200 leader failovers.
kbsync7 ms reaction, 0 idle wakeups
Swift menu bar app that keeps a MacBook's keyboard backlight in step with its screen brightness. Event-driven, with no CPU use while idle.
FaceMesh5x better than baseline
PyTorch model that animates 3D faces and turns noisy face scans into clean meshes, with USD export for Blender.
QuizzeraLive in production
Next.js and Supabase study app that writes practice quizzes from notes or PDFs with Gemini, grades them, and schedules reviews. Live at quizzera.com.
Experience
· Parallel and Distributed Computing Lab, SFU
Burnaby, BC
Jan 2027 – Apr 2027 4m
Joining Prof. Keval Vora's lab to work on distributed graph mining systems.
Oct 2026 – Feb 2027 5m Winter cohort
Migrating a website with 5,000+ users off Squarespace to a stack I run on AWS: a React frontend on S3 and CloudFront, a FastAPI backend on Lambda behind API Gateway, and PostgreSQL on RDS.
Technologies & Tools
- React
- FastAPI
- PostgreSQL
- AWS
Vancouver, BC
May 2026 – Dec 2026 8m
Built the tools engineers use to find and move 2+ PB of ultrasound inspection data, and helped run the Kubernetes and AWS platform behind them.
- Built a full-stack inspection manager with React and FastAPI, running in Docker on on-prem servers, that indexes 2+ PB of ultrasound data across those servers and AWS S3 to cut file lookups from 5 minutes to 10 seconds.
- Built a transfer runner that moves 70 TB between on-prem servers and AWS S3 with a PostgreSQL job queue.
- Sped up rendering by 30% in a C++ data server by swapping a global lock for per-dataset locks and atomics.
- Cut machine learning compute costs by 25% by auto-scaling Kubernetes GPU nodes with Karpenter and KEDA.
- Wrote reusable Terraform modules for SCPs, least-privilege IAM, and CloudTrail alarms in 10+ AWS accounts.
- Cut incident response time by 47% with Prometheus and Grafana Slack alerts across 300 Kubernetes nodes.
Technologies & Tools
- React
- FastAPI
- C++
- Docker
- AWS
- PostgreSQL
- Kubernetes
- Karpenter
- KEDA
- Terraform
- Prometheus
- Grafana
Burnaby, BC
Sep 2025 – May 2026 9m
Led 8 engineers building a volunteer scheduling app for a local animal welfare nonprofit with 1,000+ users.
- Led 8 engineers through sprint planning and code reviews to build a React app for a nonprofit's 1,000+ users.
- Built stateless FastAPI backend APIs with idempotent endpoints so retries never double-book a single volunteer.
- Cut volunteer search latency by 53% by adding MongoDB compound indexes and a Redis read-through cache.
- Cut deploy time by 60% by replacing manual server setup and releases with Terraform and GitHub Actions.
Technologies & Tools
- React
- FastAPI
- MongoDB
- Redis
- Terraform
- GitHub Actions
Burnaby, BC
Oct 2024 – May 2025 8m
Fixed a bug that caused 15% of scheduled radio shows to fail, and moved the station's website to the cloud.
- Moved the station's old PHP monolith into Docker services on AWS EKS, serving 300+ live listeners at once.
- Fixed a race condition that failed 15% of scheduled radio shows by moving jobs to asynchronous Redis queues.
- Built a Redis pub/sub and PostgreSQL pipeline that tracks listener drop-off and peak load across 50+ shows.
Technologies & Tools
- PHP
- Docker
- AWS
- Kubernetes
- Redis
- PostgreSQL
Tech Stack
Languages
Frameworks
Cloud and Infra
Tools
Point at a skill to see where I used it.
Activity
"Success is not final, failure is not fatal: It is the courage to continue that counts."