SHARKPARTY
Case studies
Case study · Platform · Full-stack · Data

RxSaver.

A prescription-discount platform—search a drug, get a coupon, save up to 80% at the pharmacy. I led its replatforming from an Azure/.NET/Angular monolith with 5-second lookups into an isomorphic React app on GCP, fed by normalized PBM data and wired for continuous experimentation.

Role—Staff Engineer & Tech Lead, RetailMeNot Live—rxsaver.com
Isomorphic ReactGraphQLKotlinPostgreSQLGCPKubernetesETLSegmentLaunchDarklyOptimizely
// The replatform

From a five-second monolith to a sub-second product.

Before
  • Staged in Azure
  • .NET back end
  • Angular front end
  • Drug lookups of 5s minimum
  • Monolithic, hard to experiment on
After
  • Cloud-native on GCP + Kubernetes
  • Isomorphic React (SSR—SEO + fast first paint)
  • GraphQL API behind a Kotlin gateway
  • Normalized PostgreSQL—sub-second lookups
  • Instrumented for continuous experimentation
// The hard part was the data

PBM feeds over FTP and email, turned into something searchable.

PBM feeds

Antiquated pricing data from Optum, Navitus, and other PBMs—arriving over FTP drops or, genuinely, emailed CSV attachments.

ETL + normalize

Multiple ETL jobs parse the inconsistent feeds and reconcile them into one normalized schema users can actually search.

PostgreSQL

A clean, indexed source of truth for drugs, pharmacies, and discount pricing.

Kotlin gateway → GraphQL

A primary Kotlin gateway fronts a GraphQL API, giving the client exactly the data it needs per view.

Isomorphic React

Server-rendered React for SEO and a fast first paint, hydrating into a rich client app.

Pharmacy benefit managers don't hand you a clean API. The pricing that powers every discount arrived as antiquated files—some via FTP, some literally emailed as CSV attachments—in formats that disagreed with each other. The unglamorous win was the ETL layer that reconciled all of it into one normalized schema, so a patient could type a drug name and get an accurate coupon in under a second.

// Experimentation & progressive UX
01
Event tracking

Segment piped a clean event stream out to the analytics and experimentation stack—one instrumentation layer, many destinations.

02
Flags & rollout

LaunchDarkly gated features and let us ship dark, roll out gradually, and kill a bad path instantly.

03
A/B & multivariate

Optimizely drove A/B and multivariate tests on the flows that mattered—search, coupon, and checkout hand-off.

04
Progressive UX

We inferred a user’s technical aptitude in real time and rendered accordingly—simpler tooling for newcomers, shortcuts and power-user affordances for the fluent.

// The point of the instrumentation wasn't dashboards—it was letting the product adapt to the person using it, in real time.

// Outcome
5s → sub-s
Drug lookups
Azure → GCP
Replatformed
SSR
Isomorphic React
3
Experimentation tools

// Honest footnote: I rolled off before the native iOS and Android apps shipped—those landed after my time on the team.

// Want the deep version?

Happy to talk through the hard parts.

Email Kam More work