Six delivered projects, documented in technical detail.

Each case study follows the same structure: the problem identified, the solution delivered, the technologies used, and the measured outcome.

Case studies

Case studies in full.

Fintech · India & GCC

Lending platform rebuilt for a 10× application volume

NBFC lending platform · 12 months · Dedicated team of 7

Challenge: A digital lending product was processing applications through a monolith that had grown for six years. Loan origination took minutes per application at peak, underwriting rules were hard-coded, and every new lending partner meant another fortnight of engineering.
Solution: We ran a four-week discovery, then peeled the origination flow out of the monolith using a strangler-fig approach. Underwriting rules moved into a configurable engine the credit team edits themselves, partner integrations moved behind a single adapter layer, and the whole pipeline became event-driven with full reconciliation.
Tech stack: JavaSpring BootPostgreSQLKafkaAWSTerraformReact

[10×]application throughput at peak

[-68%]time to onboard a new lending partner

[99.95%]platform availability over 12 months

Healthtech · India

Patient engagement app shipped across iOS and Android

Multi-city clinic network · 7 months · Fixed scope, phased

Challenge: A clinic network was running appointments across phone calls, WhatsApp and three disconnected front-desk systems. No-shows were high, records were fragmented, and patients had no way to see their own history.
Solution: We designed and built a cross-platform patient app plus a front-desk console: booking, reminders, digital records, and prescription access, with consent-based sharing and a full audit trail. Roll-out was phased clinic by clinic, with staff training at each site.
Tech stack: React NativeNode.jsPostgreSQLFHIRAzureFirebase

[-41%]appointment no-show rate

[4.6★]average store rating

[62%]bookings self-served in app

Logistics · India & Middle East

Warehouse and last-mile visibility for a 3PL operator

Third-party logistics provider · 9 months · Time & materials

Challenge: Dispatch ran on spreadsheets and driver phone calls. Customers asked where their consignment was and nobody could answer without ringing three people. Field connectivity was unreliable across large parts of the network.
Solution: We built an offline-first driver app with queued sync, a dispatch console with live map and exception alerts, and a customer tracking page. Warehouse scanning was integrated with the existing WMS, and an event stream now feeds the customer-facing ETA model.
Tech stack: FlutterGoPostgreSQLRedisKafkaGCPMapbox

[-35%]customer 'where is my order' calls

[97%]scan events captured in real time

[+22%]deliveries per driver per day

Retail & E-commerce · India

Data platform that gave one set of numbers to the whole business

D2C retail group · 6 months · Fixed scope

Challenge: Marketing, finance and operations each maintained their own version of revenue. Month-end reporting took nine working days and the numbers still had to be reconciled by hand before any board pack could go out.
Solution: We consolidated store, marketplace, ads and ERP sources into a warehouse with dbt models, added data quality tests at each layer, and built a semantic layer so every dashboard draws from the same definitions. Month-end close is now largely automated.
Tech stack: SnowflakedbtAirflowFivetranPower BIPython

[9 days → 2]month-end reporting cycle

[140+]automated data quality tests

[1]agreed definition of revenue

SaaS & Operations · UK

Document extraction assistant for an operations team

B2B services firm · 4 months · Fixed scope + support

Challenge: An operations team was keying data from thousands of supplier documents a month in five different formats. Accuracy depended on who was on shift, and the backlog grew every quarter-end.
Solution: We built an extraction pipeline combining layout parsing with an LLM step, wrapped in an evaluation harness measured against a human-labelled set. Low-confidence documents route to a review queue rather than being silently accepted, and every decision is logged.
Tech stack: PythonFastAPILLM APIsPostgres + pgvectorDockerAWS

[~80%]documents processed without human touch

[-55%]average handling time

[100%]decisions logged for audit

Manufacturing · India & EU

Legacy ERP add-ons moved off an unsupported platform

Industrial manufacturer · 11 months · Dedicated team of 5

Challenge: Fifteen years of custom modules sat on a platform that had reached end of support. A rewrite had been quoted twice and shelved twice, because nobody could afford to stop production for a cutover.
Solution: We inventoried every module, ranked them by business risk, and migrated incrementally behind a facade so both systems could run in parallel. Data was migrated with automated reconciliation at every step, and the final cutover took a single weekend.
Tech stack: .NET.NET 8SQL ServerDockerAzureAngular

[0]unplanned production downtime hours

[47]modules migrated incrementally

[-30%]annual platform running cost

About these case studies

Our policy on client evidence.

We publish what clients approve

Some of our best work sits under NDA and will never appear here. If a case study is relevant to your situation we can often walk you through it on a call.

Metrics come from your systems

Where we quote a number, it is measured in the client's own analytics or monitoring — not estimated by us afterwards.

References available

At shortlist stage we will connect you with a client working on something comparable, so you can ask the awkward questions without us in the room.

Seeking a comparable reference?

Tell us your sector and the shape of the problem. We will point you at the closest thing we have done and, where the client agrees, introduce you.

Start a project Email us directly