Cutting payment latency with a Rust-first backend pod
A growth-stage payments company needed senior systems engineers before its busiest season. A four-person pod rebuilt the authorisation hot path behind the existing gateway.
~420 → ~90 ms3 weeks
HealthtechData engineeringIllustrative composite
Rebuilding a healthtech data platform for audit-ready analytics
A diagnostics network had data spread across labs, apps and spreadsheets, and a new obligation to protect it properly. Two data engineers and a DevOps engineer built one trusted source.
3 days → same day
reporting lag for operations
1
governed definition for each core metric
Full
audit trail for access to personal data
The situation
The company runs a network of diagnostic labs and collection centres, with bookings coming through an app, partner clinics and phone. Data lived in several systems, and the weekly operations report was assembled by hand from exports. Different teams used different numbers for the same metric, and nobody fully trusted any of them.
At the same time, the leadership team wanted to be sure the company's handling of health and personal data met its obligations, including under India's Digital Personal Data Protection Act, 2023.
The brief
Build a single, trusted source for operational and business metrics, with personal data handled carefully and access that could be audited.
What we did
- Inventory and classify. Every source and field was catalogued and classified, so the team knew exactly where personal and health data lived.
- Minimise and protect. Analytics models used pseudonymised identifiers by default. Access to identifiable data was restricted to named roles, and every access was logged.
- Model the business once. A warehouse with dbt models defined each core metric in one place, with tests and documentation, so "turnaround time" meant the same thing everywhere.
- Automate and observe. Orchestrated pipelines with data-quality checks and alerts replaced manual exports.
- Infrastructure as code. The DevOps engineer built the environment in Terraform, with separate environments and reviewed changes.
How it went
Operations moved from a weekly, hand-assembled report to dashboards refreshed the same day. Arguments about whose number was right largely stopped, because there was now one governed definition for each core metric. The company could show who had accessed personal data, when and why.
What we'd tell another team
- Start with the inventory. You cannot protect or model data you have not found.
- Agree metric definitions with the business before writing pipelines.
- Privacy by default is cheaper than privacy retrofitted.
Data engineering is mostly agreement engineering. The people we place are as good in the definitions meeting as they are in the pipeline code.
Saurav Kumar Jha, Founder & CEO
More case studies
All case studies Other ways engagements run.
Illustrative case
Illustrative caseStanding up an AI support team for a D2C brand
A fast-growing consumer brand was drowning in support tickets every sale season. Two engineers built a retrieval and agent system with an evaluation harness and a human in the loop.
~45%Hours → minutes
Illustrative caseBuilding an outbound engine for a SaaS company entering the US
An Indian B2B SaaS company with strong local traction wanted to open the US mid-market. A business development manager and a GTM engineer, hired together, built the motion from positioning up.
123 sequences
Hire with Engagetal
Your next great hire is already vetted.
Tell us the role in five questions. We come back with four profiles who have cleared every stage of our screening, and you only pay when someone joins.