Software Engineer · BRAC IT Services Ltd.

Backend engineering for systems that cannot afford ambiguous behavior.

I build CRM, ticketing, and lending services with Java and Spring Boot. My work focuses on configurable business rules, workflow automation, data consistency, and reliable messaging, with independent labs exploring concurrency, security, and failure handling.

Professional experience at a glance

Three production domains. One focus on correctness.

Building backend services at BRAC IT Services Ltd. since December 2024, across CRM, ticketing, and loan-proposal workflows. See complete experience ↗

01 · CRM

Rule-driven sales operations

Lead-to-quote workflows, configurable assignment rules, idempotent Kanban moves, and traceable stage transitions.

02 · Ticketing

Least-loaded resolver routing

Redis-backed workload counters, min-heap assignment, notifications, and SLA/KPI tracking.

03 · Lending

DDD/CQRS loan proposals

Command-side microservice with 21 composable business specifications, bulk workflows, and lifecycle domain events.

Flagship engineering case study

Spring Boot Rescue Lab

A deliberately fragile order API repaired one production-style incident at a time. Instead of presenting a clean CRUD demo, the repository preserves the failure, root cause, trade-offs, remediation, regression test, and evidence for each incident.

42 → 3 SQL statements
8 → 1 concurrent retry outcome
2 → 1 buyers accepted for last unit
401 / 403 / 404 / 200 authorization matrix
At-least-once reliable event delivery model
24 → 1 cold-cache PostgreSQL loads
INC-001

SQL performance

N+1 order search

A 20-order page generated 42 prepared statements because DTO mapping traversed lazy order items and products after pagination.

Fix
Pagination-safe ID page + graph fetch + composite PostgreSQL index.
Evidence
42 → 3 statements in the controlled 20-order test; 92.86% fewer SQL statements.
Read incident report ↗
INC-002

Idempotency

Duplicate orders after retries

A timed-out client could retry a successful request and create a second order because the API had no durable command identity.

Fix
Customer-scoped idempotency key, SHA-256 fingerprint, advisory lock, durable replay record.
Evidence
8 concurrent retries → 1 order, 1 stock reservation, 7 replayed responses.
Read incident report ↗
INC-003

Concurrency control

Inventory overselling

Two buyers read the same final unit and both succeeded through a classic lost-update race, even though the final stock value still looked valid.

Fix
Atomic conditional stock update inside the order transaction.
Evidence
2 buyers / 1 unit → exactly 1 success and 1 insufficient-stock rejection.
Read incident report ↗
INC-004

Application security

Broken object authorization

The baseline trusted a caller-controlled customer header and allowed arbitrary order access, administrative actions, and operational endpoints.

Fix
Principal-derived identity, ownership-scoped queries, explicit role boundaries, deny-by-default security.
Evidence
Real Spring Security integration matrix covering anonymous, customer, cross-customer, and admin paths.
Read incident report ↗
INC-005

Messaging reliability

Lost events across PostgreSQL and RabbitMQ

A database commit and broker publish are separate side effects. A crash between them could leave a durable order with no integration event for downstream systems.

Fix
Transactional outbox, publisher confirms, retries with backoff, DLQ, and consumer deduplication.
Guarantee
Committed orders retain durable event intent; duplicate delivery is expected and harmless under an explicit at-least-once model.
Read incident report ↗
INC-006

Cache resilience

Hot product cache stampede

A naive Redis cache still allowed every request to hit PostgreSQL when a popular key expired, multiplying database work during the exact burst the cache was meant to absorb.

Fix
Redis distributed single-flight lock, double-checking, TTL jitter, bounded waiting, and after-commit invalidation.
Evidence
24 simultaneous cold reads → exactly 1 PostgreSQL load; Redis outage fails open to the database.
Read incident report ↗

Controlled lab evidence, not production traffic: deterministic reproduction · root-cause analysis · alternatives considered · Flyway migrations · PostgreSQL + Redis Testcontainers · JUnit 5 · CI evidence artifacts · failure injection · correlation IDs · ECS JSON logs · Actuator · Prometheus · Grafana

Professional experience

Backend systems with real operational workflows.

Software Engineer at BRAC IT Services Ltd. · December 2024–present · Dhaka, Bangladesh. Building business workflows across CRM, ticketing, and loan proposal services.

BRAC IT Services · CRM platform

Rule-driven sales operations

Developed Spring Boot services for lead-to-quote workflows and sales pipelines; implemented configurable ownership assignment based on priority, schedules, conditions, and rotation rules.

  • Optimized Kanban ordering by replacing full-column reordering with fixed-length order keys, reducing routine card moves to a single rank assignment.
  • Implemented idempotent card moves, board access controls, and audit trails; synchronized opportunity columns with pipeline stages and validated allowed transitions before changing business state.

Spring Boot · PostgreSQL · Flyway · Redis · RabbitMQ · OpenAPI

BRAC IT Services · Ticketing platform

Support operations, routing, and SLA visibility

Developed ticket lifecycle services covering creation, search, comments, forwarding, assignment, and notifications using Spring Boot, PostgreSQL, Redis, and RabbitMQ.

  • Automated least-loaded ticket assignment using Redis-backed workload counters and a min-heap, incorporating resolver availability and reliever fallback to reduce manual routing.
  • Implemented SLA and KPI tracking with assignee-restricted stage updates, progress notifications, and activity histories; maintained workload counters through asynchronous updates following ticket creation, reassignment, and resolution.

Spring Boot · PostgreSQL · Redis · RabbitMQ · JPA/Hibernate

BRAC IT Services · Lending services

Loan proposals with explicit business rules

Developed command workflows for loan proposal creation, updates, bulk processing, approval, and soft deletion in a DDD/CQRS microservice using Spring Boot, MongoDB, and RabbitMQ.

  • Encapsulated 21 composable business specifications covering eligibility, loan limits, installments, insurance, nominees, and disbursement within a loan-proposal aggregate; added aggregate-level tests for proposal updates.
  • Implemented atomic monthly proposal-number generation, batch duplicate checks, and lifecycle domain events to support proposal identification, validation, and downstream integration.

Spring Boot · MongoDB · RabbitMQ · DDD · CQRS

Selected systems

Focused builds for distributed backend problems.

I use independent projects to make architecture choices inspectable and to explore failure modes that are difficult to demonstrate safely in production systems.

Identity and access management · Independent lab

Enterprise IAM Lab

Two OIDC clients demonstrate Authorization Code + PKCE, shared SSO sessions, LDAP user federation, and role/group mapping. A Spring Boot resource server validates JWT issuer and audience, maps Keycloak roles to Spring authorities, and enforces method-level access for employees, support agents, and administrators.

KeycloakSpring SecurityOAuth 2.0 / OIDCPKCEOpenLDAP
View repository ↗

Distributed transactions · Proof of concept

Spring Saga Orchestrator

Event-driven order coordination across order, inventory, and payment services with separate data ownership, event-ID tracking, explicit failure events, and compensation paths.

Java 25Spring Boot 4RabbitMQMongoDB
View repository ↗

Event-driven delivery

NotifyFlow

Notification delivery and latency simulation with durable delivery records, asynchronous fan-out through RabbitMQ, and live per-user status updates over WebSocket/STOMP.

JavaSpring BootRabbitMQPostgreSQL
View repository ↗

Capabilities

What I optimize for.

Framework knowledge matters, but the harder part is designing behavior that stays correct when requests repeat, transactions overlap, dependencies fail, and systems evolve.

01

Correctness before scale

Define invariants, transactional boundaries, ownership, and failure semantics before optimizing throughput.

02

Evidence before optimization

Reproduce the failure, measure it, compare alternatives, and keep regression evidence close to the code.

03

Operations are part of design

Retries, queues, backlogs, logs, metrics, health checks, and recovery paths are product behavior.

Backend

Java 25 · Spring Boot 4 · Spring Security · Spring Data JPA · Hibernate · REST · OpenAPI

Architecture

DDD · CQRS · Saga · modular monoliths · microservices · transactional outbox · idempotency

Data & messaging

PostgreSQL · MongoDB · Redis · RabbitMQ · Flyway · indexing · query design

Testing & delivery

JUnit 5 · Mockito · Testcontainers · GitHub Actions · Gradle · Docker · Docker Compose · Git · Linux

Identity & security

Keycloak · OAuth 2.0 · OpenID Connect · PKCE · JWT · role-based access control

Observability

Spring Boot Actuator · Micrometer · Prometheus · Grafana · structured logging

Background

Engineering foundation.

Computer-science training, production backend ownership, and research discipline.

Dec 2024 — Present

Software Engineer

BRAC IT Services Ltd. · Dhaka, Bangladesh

Backend engineering for CRM, ticketing, and DDD/CQRS loan proposal services.

2024

B.Sc. in Computer Science & Engineering

Chittagong University of Engineering and Technology (CUET)

CGPA 3.59 / 4.00

2023–2024

Competitive programming achievements

Codeforces Specialist (1500+ rating); CUET IUPC finalist in 2023 and 2024; CUET Tech Carnival 2020 finalist.

Contact

Looking for a backend engineer who can reason about failure?

I am interested in Java/Spring backend roles involving distributed workflows, reliability, data consistency, security, and production-scale systems.