AWS · iGaming · Sports Betting
15M+ daily requests, sub-100ms latency — a betting platform re-engineered for peak.
We rebuilt a tier-1 betting operator's live wagering stack on a multi-region, event-driven architecture — autoscaling odds ingestion, in-memory bet validation and a hardened edge to absorb match-day surges without a single dropped ticket.
[Requests served]
15M+/day
[P95 bet latency]
<100ms
[Infra cost]
-46%
[Match-day uptime]
100%
[ The Problem ]
Where the business was stuck.
Legacy wagering stack couldn't handle match-day traffic surges — dropped tickets, slow odds updates and spiralling infra costs during peak events.
[ Key Challenges ]
- ▸Sub-second odds updates across millions of concurrent users
- ▸Zero-tolerance for dropped bets during match-day spikes
- ▸Runaway infra cost from over-provisioned peak headroom
Our approach
How we engineered
the outcome.
STEP 01
Multi-region active-active architecture on AWS with Route 53 latency routing
STEP 02
Event-driven odds ingestion on MSK (Kafka) + ElastiCache for in-memory bet validation
STEP 03
Autoscaling on custom demand signals — not just CPU — using KEDA on EKS
STEP 04
Progressive delivery of pricing engine changes with instant rollback
[ Solution highlights ]
Delivered — measured — in production.
- ✓Odds pipeline latency dropped from 480ms to <90ms P95
- ✓Bet validation moved fully in-memory with Redis + DynamoDB streams
- ✓Edge WAF and rate-shaping absorbed 3× match-day spikes
- ✓Cost model shifted to spot + graviton for 46% infra savings
[ Tech stack ]
Built on AWS.
AWS EKS
MSK / Kafka
ElastiCache
DynamoDB
Route 53
KEDA
Graviton
"
Match-day used to be war-room mode. Now it's a non-event — the platform scales itself and we watch the game.
Keep exploring