Data Stories

We show the work.

Build stories from our own products. Benchmarks we ran ourselves. Blueprints for public problems. Real client work. Each labeled as exactly what it is — no anonymous “Fortune 500 client” fiction here.

Build storySomething we built. Architecture, decisions, tested numbers.
BenchmarkTests we ran ourselves. Numbers, method, repo.
BlueprintA public problem, solutioned in full. How we'd build it.
Client storyReal delivered work.
Build story In the lab — Aug 2026

Two books, one truth: inside Pravah's co-lending reconciliation engine

Bank ledger vs NBFC ledger, tolerance rules, a named break taxonomy — and what happened when we threw millions of synthetic transactions at it.

PravahReconciliationNBFC
Email me when it ships
Benchmark In the lab — Aug 2026

dbt incremental strategies on Databricks: merge vs insert_overwrite vs microbatch

Same data, three strategies, real costs — including dbt's new microbatch, and which strategies stay correct when yesterday's data arrives today.

dbtDatabricksBenchmark
Email me when it ships
Build story In the lab — Sep 2026

A lakehouse where data can't leave India: DriveLakeDB on Iceberg + Trino

Residency-first architecture, DPDP alignment as code, and what breaks when you refuse the US-region shortcut.

IcebergTrinoDPDP
Email me when it ships
Benchmark In the lab — Sep 2026

The real AWS bill of a badly partitioned table

Same queries, same data, three partitioning choices — and the monthly cost delta, measured, not estimated.

AWSAthenaCost
Email me when it ships
Build story In the lab — Oct 2026

GSTR-2B at scale: why CA firms drown in reconciliation, and the rail we built

Pravah's first module: the ITC-matching problem, a named mismatch taxonomy, and the human-confirm loop that keeps the signature with the professional.

PravahGSTCA firms
Email me when it ships

Reading this because you have one of these problems?

Skip the form. One call, and you leave with next steps — whether or not you hire us.

Book a data-readiness call