12 Hours → 30 Seconds
Founder Context
Mid-size asset manager · ~$15B AUM · CTO-led, internal team capacity-constrained
How we rebuilt a mid-size asset manager's risk analytics from overnight batch jobs to live real-time, in 16 weeks — with zero downtime during the cutover.
The Founder Situation
A mid-size asset manager had a risk-analytics problem costing them deals. Their existing platform calculated risk overnight in batch jobs — by morning, the numbers were stale. Competitors had moved to real-time monitoring. Their largest institutional clients were starting to notice. The internal CTO had a clear technical vision and a small team that couldn't take on the rebuild without dropping their compliance roadmap. They needed an external squad to own the latency-critical path for one focused build, ship it, and hand it back.
Why We Were a Fit
Not a long-term engineering partnership. A focused 16-week build with a non-negotiable Q-end deadline tied to a client commitment. Internal CTO involved on architecture; external team owns the build, the cutover plan, and the parity proof.
How We Shipped It
- Co-designed the streaming architecture with the internal CTO in the Roadmap phase: Kafka for events, ClickHouse for time-series, Go for the calculation engine, Next.js for the dashboard
- Working demos every Friday with the CTO and the head of trading from week 4 onward
- Parallel-run period weeks 10–13: new platform ran in shadow mode against the legacy batch system, differences flagged and investigated daily
- AI-paired delivery on schema migrations, test fixtures, and dashboard plumbing; senior judgment on the streaming architecture, the cutover-risk plan, and the parity-check methodology
- Production cutover weeks 13–15 with monitoring stand-up and on-call runbooks for the internal team
What We Built
A 3.5-person team — one senior engineer with 12+ years in financial systems and distributed real-time architectures, one builder paired with AI tooling, one fractional designer for the trading-floor dashboard, and the studio founder as product partner. Live trading data ingested via Kafka, calculations in Go, time-series in ClickHouse, sub-30-second risk recalculation, real-time dashboard built on Next.js. Zero downtime cutover. Clean handover to the internal team at week 16.
What This Means for Your MVP
If your MVP touches real-time data, financial calculations, or any latency-critical surface, this case is the closest analog. Latency requirements force honest architecture decisions. Boring technology, unconventionally combined. AI on the boilerplate, senior judgment on the parts that matter. Parallel-run before cutover. Clean handover at the end. The model shape works.
- 12-hour batch jobs collapsed to sub-30-second realtime calculations
- Zero downtime during the production cutover
- Provable parity with the legacy system before cutover (3-week parallel-run period)
- Internal team owned operations from day 112
- 16-week timeline hit on the original Q-end commitment date
Have an idea with a similar shape?
30 minutes with the studio founder. We'll tell you honestly whether your MVP fits the model.