Nuro
Each of these is a real decision Nuro is working through, written so you can make the call yourself — incomplete information, someone who pushes back, no answer key.
Nuro is an autonomous-driving technology company that develops and licenses the Nuro Driver, an integrated Level 4 system of AI driving software, vehicle hardware, sensors, onboard computing, and remote-assistance tools. It takes a machine-learning-first approach to autonomous driving technology. It serves automakers, mobility providers, ride-hailing platforms, and commercial-fleet operators through B2B licensing and technology integration. Nuro is expanding across new geographies throughout the U.S. and globally and across new vehicle platforms including robotaxi, personally owned vehicles, and delivery/logistics. It has a ride-hailing product that runs on a real-time distributed system. It maintains an in-house ML compiler platform that compiles, deploys, and validates its autonomy models on various hardware platforms, and it validates every autonomy code change for real-time performance on actual robot compute hardware before the change reaches the road.
Your first one is free. No card.
27 live right now.
The rest
Package incidents into replays without losing the software version
Incident reviewers cannot connect video, telemetry, and the driving software into one replayable event. You have to weigh fast evidence assembly against traceability and dependable simulator context.
Choose a safe stop before the vehicle window closes
Some emergency-scene cases enter safe-stop too late, while others create avoidable remote-assistance calls. You must weigh responder safety against a fallback behavior that could constrain planned driverless operations.
Fix stream order after short reconnects without loosening rate control
Priority vehicle streams are resuming behind secondary views after brief reconnects. You must weigh faster recovery against stable rate control for every active stream.
Confirm operator readiness before the weekly allocation is published
New-capability driving hours are being reassigned after operators reach the shift without consistent preparation. You must weigh predictable allocation against a readiness process operators will trust enough to use.
Set safer shifts while validation coverage is already committed
Several late and overnight operators report fading attention near the end of long driving shifts. You must weigh route coverage against a shift pattern that operators and safety reviewers can defend.
Recover overnight routes when shift prep consumes the driving window
Overnight routes are leaving late even when operators arrive on time, cutting planned driving hours. You must weigh a faster handoff against the vehicle checks that keep every drive safe.
Keep sensing clear through spray without adding depot stops
Wet-route vehicles need manual sensor cleaning between runs, which makes availability hard to plan. You must weigh weather resilience against package space, serviceability, supplier readiness, and pilot-build cost.
Choose a weather seal while the pilot build window closes
Sensor resets after rain exposure could remove the remaining pilot-build path. You have to weigh a tooling commitment against evidence that still supports more than one failure mechanism.
Secure radar mounts while the next vehicle build stays on schedule
Radar assemblies are dropping availability after thermal cycling, even though they pass initial checks. You must weigh a durable mechanical correction against the time and packaging limits of the next build.
Decide who can see a vehicle trace before the first deployment review
Diagnostic access is still handled through private handoffs, leaving no clear record of who saw a vehicle trace. You must weigh investigation speed against access that remains narrow and auditable.
Choose a release path without removing the stalled-vehicle protection
Remote-assistance operators are seeing more vehicles pause near emergency scenes after the new software version. You have to weigh a fleet halt against removing a safety protection while the evidence remains incomplete.
Merge cases from one reconnect without hiding separate losses
One communication loss can open several investigations after a vehicle reconnects. You have to reduce duplicate work without hiding a separate fleet event.
Trace release evidence before operators review first routes
Operator reviews rely on screenshots that do not reliably tie evaluation results to one software version. You must weigh a useful first evidence dataset against the time needed to validate its lineage.
Settle which scenes belong in one comparison before evidence locks
A planning gain appears in the dashboard, but rare scenes entered only one comparison cohort. You must weigh a five-day release decision against evidence that stays defensible after deployment.
Stop retries from duplicating records while daily refreshes continue
Duplicate retries are making rare behavior coverage appear larger than it is. You must weigh fast stream recovery against evidence reviewers can trust.
Reconcile Jira and Confluence before leaders commit launch dates
Conflicting Jira and Confluence updates leave robotaxi launch dependencies with unclear owners and dates. You must weigh fast status visibility against the human judgment required before a critical-path record changes.
Decide what blocks launch in responder cases before paid rides begin
The partner cannot yet trace disabled-vehicle and responder scenarios to one approved readiness record. You have to weigh a launch bar that is credible to reviewers against one the first market can actually meet.
Commit the fleet release while construction evidence still conflicts
Temporary traffic-control cases are producing conflicting simulation and vehicle-validation evidence ahead of a fleet software lock. You must weigh the partner’s deployment sequence against the proof required for driverless readiness.
Fix the fix-to-dashboard handoff so closed failures stop showing as open
Validated fixes remain visible as open risks and pull engineers into duplicate investigation. You must weigh a faster status handoff against preserving an auditable evidence trail.
Rank new routes by risk until offline evidence earns trust
Program managers lack a reliable way to find the route cohorts most likely to regress after expansion. You must weigh a useful early ranking against the risk of overstating what sparse history can predict.
Decide on emergency scenes while the evidence conflicts
Simulation shows a 3.6% late-lane-change rate, while 19 on-road traces remain inconclusive. You must weigh a locked fleet branch against evidence that is expensive to verify.
Stabilize wet-road mapping before the next model release
Wet-route boundary inconsistency rose from 1.1% to 8.4% after the latest model update. You must weigh a fast model-side correction against preserving the aggregate gain.
Triage failed benchmarks without weakening the release gate
Known benchmark failures are consuming validation time before engineers can decide whether to rerun or intervene. You must weigh faster evidence gathering against the review boundary that protects release-blocking validation.
Prove releases on Thor benches without taking NPI hardware windows
Vehicle-platform teams cannot reliably connect a planned release to one comparable NVIDIA Thor result. You must weigh faster self-service access against protected hardware capacity and trusted evidence.
Choose the latency gate while the bench fleet disagrees
Mixed NVIDIA Thor results are blocking a release during a shrinking validation window. You must weigh release momentum against evidence that the gate still reflects vehicle behavior.
Clear completed benchmarks from the gate without loosening validation
Completed NVIDIA Thor runs are leaving autonomy changes blocked after their results reach BigQuery. You must weigh faster merge decisions against preserving trustworthy release evidence.