Work that ships — boring systems, loud outcomes.

Four anonymised case studies from engagements in the last 18 months. Full results decks, client logos, and NPS breakdowns are available under NDA on request.

Fintech · Payments 2025
Client
Series B payments platform, $80M TPV
#Postgres#Kafka#Idempotency#Multi-region

Payments infra — 99.999% uptime target

Led a 10-week architecture review of the core ledger and checkout pipeline. The system was shedding 400+ sessions/day during US East peak and had rolled back 3× in the prior quarter.

Problems we had to solve
  • Distributed transactions spanning ledger and downstream settlement with no idempotency keys
  • Cross-region failover broke sequence-generator uniqueness guarantees
  • Monitoring stack had 3 sources of truth — no single alerts dashboard for payment outcomes
Stack & approach
  • Postgres 16 · logical replication
  • Kafka · exactly-once Streams topology
  • Go idempotency middleware
  • Grafana + single SLO dashboard
$0.8M
Incident-related annual spend eliminated
42%
p99 checkout latency reduction
30%
Postgres write I/O reduction via idempotency layer + batching
0
Rollbacks in 2 quarters following rollout
Developer Tools · AI 2025
Client
Seed → Series A AI code-collab tool, 120k signups
#DevRel#Pricing#Positioning#Messaging

AI dev-tool PLG funnel redesign

Hired to diagnose a leaky PLG funnel. Signups were growing 10% month/month but activation was stuck. Founder had burned 2 GTM agencies that “just didn’t get devs.”

Problems we had to solve
  • Hero message was “code faster with AI” — identical to 27 competitors on ProductHunt
  • Onboarding tour showed value after 17 clicks; 71% dropped on step 3
  • Pricing used per-seat pro for 3 tiers with no solo/hobby hook — 83% of activated users were on the free tier forever
Stack & approach
  • Positioning matrix vs 14 competitors
  • Hero + landing-page rewrite
  • 3-tier pricing with Solo hook
  • Developer-first announcement thread + Show HN playbook
61%
Activation (was 31%) — 10 weeks
4.2×
Paid seat volume
2.3×
Mean paid-30-day cohort value
6 mo
Series-A closing accelerated by 6 months
SaaS · Multi-tenant 2024
Client
B2B SaaS analytics, 3k paying customers, 4 regions
#CDN#Edge#Caching#Observability

Multi-region SaaS perf overhaul

Customer surveys were littered with “slow” / “dashboard is broken” tickets. Enterprise prospects were ghosting the demo step after seeing EU-region LCP in Gtmetrix.

Problems we had to solve
  • Every tenant had its own origin hit on every request — no CDN strategy beyond static assets
  • SSO + dashboard bootstrap did 9 sequential RTTs in the worst case
  • Postgres read replicas were lagging 2–3s during EU morning peak
Stack & approach
  • Cache strategy: signed URLs + short TTL + stale-while-revalidate
  • Edge-rendered critical CSS + fonts
  • Postgres connection pooling + PgBouncer
  • Grafana-Faro RUM across 4 regions
1.1s
Global 75th-percentile LCP (was 3.8s)
68%
API p95 latency reduction
+14pp
Enterprise 90-day retention
37%
Origin bandwidth cost reduction
B2B SaaS · Growth 2024–2025
Client
Bootstrapped B2B SaaS, 8 figures ARR
#Hiring#Roadmap#Architecture#Platform

Advisory CTO — scaling from 6 engineers to 22

Founder-CTO was moving to CEO. Engineering had 6 people, a monolith that hadn’t been split, and a single 4Q platform roadmap they’d missed 3 times running.

Problems we had to solve
  • No hiring rubric or calibration — 3 recent senior hires missed expectations badly
  • Platform roadmap lacked OKRs and was tied to one risky rewrite
  • Deploys to prod were manual Friday rituals with 2 engineers on a checklist
Stack & approach
  • Hiring scorecards + 4-step calibration loop
  • Strangler-fig platform split plan
  • GitHub Actions + ArgoCD blue/green deploys
  • Quarterly OKRs with 2-way roadmap sync
8
Senior engineers hired and ramped on-track
3 mo
Platform rewrite completed early
12/min
Deploys to production (was 0.3/day)
38%
Median time-to-first-PR reduction for new hires