Hot Takes. Cold Beers. How to evolve high-scale backend systems without breaking production.
Most backend systems don't fail all at once. They slow down. They accumulate debt. A JOIN here, a replicated data store there, an on-call runbook that quietly doubles in length every quarter — until one day, you're staring down 400 million daily requests on a single endpoint, or watching a production incident drag three engineering teams into a six-hour recovery spiral.
This evening is for engineers who've been there — or can see it coming.
Join us the night before WeAreDevelopers for cold drinks, Berlin sausages, and two of the most honest production stories you'll hear all week. No ideal-world architecture diagrams. No vendor pitches. Just the real decisions, real trade-offs, and real results from the GetYourGuide engineers who had to fix things while the plane was still flying.
Come for the hot takes. Stay for the cold drinks.
Agenda:
17:30 to 18:30 Networking — cold beers, sausages, good conversations
18:45 Welcome from GetYourGuide Engineering
19:00 🔥 Talk 1 — Re-designing reviews read path for high scale and performance
19:45 Break
20:00 🔥 Talk 2 — Untangling a Distributed Monolith
21:00 End
The Talks
Talk 1 — Re-designing reviews read path for high scale and performance: How we refactored GetYourGuide's review read path architecture without breaking production
with Primanshu Choudhary
User-generated reviews are the trust signal at the heart of GetYourGuide's marketplace. But what happens when the service powering them starts to buckle under its own weight?
In this talk, Primanshu will walk through how he tackled a real bottleneck in GetYourGuide's reviews service: a monolithic reviews endpoint handling 350M+ requests per day that had accumulated years of tech debt, inefficient database patterns, and responsibility overload — leading to latency regressions, SLO breaches, and a growing on-call burden.
The hot take: You don't need a big-bang rewrite. They'll share the concrete decisions that turned things around:
Eliminating 23+ complex JOINs through strategic denormalisation of the hot fields
Scaling down from 4 reader nodes to 2 — and eliminating a DB caching layer entirely
Killing expensive COUNT(*) aggregations with smarter pagination
Reviews endpoint: ~30% p95 latency improvement; p99 dropped 115ms to 45ms (60% faster) on the slowest 1% of requests
Reduced DB load which helped us save thousands of dollars in cost (~$70k annually).
All of it shipped with zero service or database downtime
Whether you're running a read-heavy system under load or just starting to feel the creep of database strain, this talk delivers a pragmatic playbook for serving millions of requests at low latency — without blowing everything up to do it.
Talk 2 — When multiple services become one: Untangling a distributed monolith
with Ayush Kumar
At GetYourGuide, a few of our core backend services — Connectivity, Inventory, Catalog — quietly evolved into a distributed monolith over years of organic growth. Neither service could function without the others. Business data was replicated across four separate data stores. Recovering from a production incident meant coordinating deployments across multiple engineering teams.
When this architectural coupling started causing customer-facing failures — with recovery times reaching 8–10 hours — a cross-functional team of four engineers paused feature work for four weeks to redesign the architecture from first principles.
The hot take: The warning signs were there 18 months earlier. They ignored them. Here's what that cost.
Ayush shares the complete production story:
The five specific architectural decisions that eliminated the coupling
A zero-downtime migration strategy for 2TB of live data
How recovery time dropped from 6 hours to 60 seconds — a 98% reduction
Deployment coordination cut from 3 teams to 1
The business case that convinced leadership to approve a four-week task force
And the honest trade-offs: +40ms read latency, careful migration planning, and what they'd do differently
Real scale. Specific patterns. Measurable outcomes. No sugarcoating.
Bonus: Skip the morning queue 🎟️
GetYourGuide is an official WeAreDevelopers pre-check-in point. Grab your conference badge on your way into our event and beat the lines on the day of the event.
About GetYourGuide
GetYourGuide is a leading global online marketplace to discover and book experiences worth traveling for. Travelers can use GetYourGuide to find things to do in more than 18,000 cities, including tours from local experts, exclusive access to must-see attractions, as well as immersive bucket-list experiences. More than 50,000 supply partners leverage GetYourGuide’s easy-to-use platform to grow their businesses, offering 200,000 experiences to travelers around the world. For more information, follow GetYourGuide on LinkedIn, Instagram and TikTok and visit getyourguide.com.