DEV3Technologies
All case studies

FinEdge Capital

Month-end close from three days to four hours with a distributed queue

Batch reconciliation scripts replaced with idempotent, partitioned workers on a distributed queue — with a regulator-ready audit trail for every ledger movement.

Node.jsPostgreSQLDistributed QueueRedisKafkaKubernetesAWS
Industry
Fintech
Region
India
Duration
18 weeks
Team
5 engineers
Services
Custom SoftwareCustom SaaS Engineering

Impact

Month-end close

4 h94%

Before: 72 h

Auto-matched transactions

96%57%

Before: 61%

Jobs processed / minute

18,00015×

Before: 1,200

Live project preview

Client systems are private — screenshots shown where permitted.

FinEdge Capital — preview under NDA

Production screens available under NDA on request

The challenge

Reconciliation ran as nightly cron scripts against a single PostgreSQL instance. A failure halfway through meant restarting from scratch, month-end close took three days of analyst time, and auditors could not trace how a balance had been derived.

Our solution

Partitioned, idempotent workers

Statements are split into account partitions and processed by BullMQ workers backed by Redis, with Kafka carrying settlement events from upstream banks.

  • Each job is idempotent and resumable, so retries never double-post.
  • Back-pressure and per-partner rate limits protect bank APIs.
  • Dead-letter queues route exceptions to an analyst review screen.

Immutable audit trail

Every ledger movement records its source statement, rule version and operator, queryable from the finance console.

Results

  • Month-end close reduced from 72 hours to 4 hours.
  • Auto-matched transactions rose from 61% to 96%.
  • The regulator accepted the audit export without follow-up requests.

Related case studies