Work we're proud of.

Scaling a Tier-1 European DSP bidding engine to 180k RPS

The challenge

A leading European demand-side platform was hitting throughput ceilings during peak auction windows. Their Go-based bidding service degraded above 80k RPS, causing auction timeouts and lost bid opportunities worth millions per month.

Our approach

We profiled the hot path, redesigned the in-process cache layer with lock-free ring buffers, switched from HTTP/1.1 to gRPC for internal service communication, and introduced back-pressure signalling to the auction router.

Tech stack

Snowflake migration: 60% cost reduction in 8 weeks

The challenge

A large Polish e-commerce retailer was running an aging Redshift cluster with a tangled spaghetti of SQL transformations. Query times were growing, costs were unpredictable, and the data team couldn't iterate on analytics models without breaking production pipelines.

Our approach

We migrated the warehouse to Snowflake, rebuilt the transformation layer with dbt (400+ models, full lineage), introduced Airflow for orchestration and established a DataOps CI/CD pipeline with automated testing and data quality checks.

Tech stack

Zero-downtime event ingestion at 2.5M events/hour

The challenge

AdShift needed a production-grade event ingestion pipeline capable of handling mobile attribution events at scale, with strict latency requirements and full auditability for GDPR compliance.

Our approach

We designed and shipped a Go-based ingestion service on EKS with Kafka for durable buffering, External Secrets for secret rotation without restarts, and a Grafana/OpenTelemetry observability stack. Blue-green deployments via ArgoCD ensured zero-downtime releases.

Tech stack