booting the lab environment
0%
live environment · no slides, real systems

The Lab.

Five years of building data systems, turned into things you can actually play with. Assemble a pipeline, survive a 2am on-call shift, tune a Spark cluster, and query a live warehouse. Everything here is hand-built and runs in your browser.

ops-monitor · prod
Throughput
0
p99 latency
187ms
Throughput stream
Game 01 · Build the pipeline

Put the stages in order.

Tap the stages to drop them into the slots, source to serving. Get the flow right and the pipeline goes live. Tap a filled slot to send it back.

Pipeline builder
Drop all six stages to check the flow.
Game 02 · On-call simulator

It's 2am. Keep it alive.

Alerts fire faster than you'd like. Resolve each one before its timer runs out, or system health takes the hit. Thirty seconds on the pager. How high can your uptime stay?

Incident response
System health
Score0
Time30
Best0

Ready for your shift?

Incidents will appear below. Click resolve before each timer empties. Survive thirty seconds.

Game 03 · Spark tuning

Find the sweet spot.

More executors is not always the answer. Balance parallelism, partitioning, and batch size to hit the throughput goal at the lowest cost. Watch the numbers move as you drag.

Cluster simulator
Executors8
Shuffle partitions64
Batch size (MB)64
Throughput
0
p99 latency
0
Cluster cost
0
Efficiency
0
Goal: 80k records/s under 200ms p99, as cheap as possible.
Game 04 · Warehouse console

Query the warehouse.

A real prompt wired to a small in-browser warehouse. Run queries against my own data. Type help to see what is there, and look for the easter eggs.

phaneendra@warehouse
phaneendra@warehouse:~$
Try help · show tables · select * from experience · whoami · hire
That's the lab

Built the way I build everything.

Every game on this page is hand-written, no libraries, no shortcuts. The same care goes into the pipelines I ship. If that is the kind of engineer your team needs, let's talk.