Open to Software / Backend / DevOps / AI Engineering conversations
Badda, Dhaka, BangladeshAsia/Dhaka (UTC+6)
Md. Musfiqur
Rahaman
I work as aFull Stack Software Engineer
Roles: Full Stack Software Engineer, Backend Engineer, AI Engineer, DevOps Engineer, Software Engineer.
I design the systems underneath the product - multi-tenant SaaS platforms, distributed crawlers, streaming pipelines and the AI layers on top of them - and I stay responsible for them all the way to the container running in production.
- 3+
- years shipping production systems
- 10k+
- daily active users supported
- 4
- platforms designed end to end
- 5
- engineering disciplines covered
- Full StackDjango · DRF · FastAPI
- BackendPython · Django · DRF
- AILangChain · LangGraph · CrewAI
- DevOpsDocker · GitHub Actions · Traefik
- EngineeringPython · TypeScript · C++
Core stack: Python, Django, FastAPI, PostgreSQL, Redis Streams, Celery, pgvector, Docker, Traefik, AWS, Next.js, React Native, LangGraph, Scrapy, Playwright, Prometheus, Grafana, Cloudflare, PgBouncer, WebSocket.
01 / Capability console
Five roles, one engineer
Job titles slice the same work differently. Pick the lens you are hiring for - the diagram, the evidence and the stack all re-focus around it.
Hover a nodeEach block is something I have built, broken and fixed in production.
One person, the whole vertical slice.
I take a feature from an ambiguous requirement to a deployed, observable slice - schema, API, UI state, background jobs and the release that carries it. Three products in production, two of them multi-tenant SaaS with a web app and a mobile client on the same API.
- 3production platforms owned end to end
- Web + Mobileclients served from one Django API
- 10k+daily active users supported
- Django
- DRF
- FastAPI
- React
- Next.js
- React Native
- TypeScript
- PostgreSQL
- WebSocket
- 01
Product-shaped API design
RESTful resources modelled on the domain, not the ORM - versioned, paginated, permission-aware, and documented well enough that the frontend team never has to read my code.
- 02
React & Next.js front ends on my own contracts
Dashboards, onboarding flows and admin surfaces in React, Next.js and React Native, wired to WebSocket channels for live state instead of polling.
- 03
Requirement → architecture → delivery
I sit in the requirements conversation, translate it into a data model and a delivery plan, then own it end to end. That loop is why I lead junior engineers today.
Correctness under concurrency is the whole job.
Most of my hardest work lives below the API: lease-based work queues, at-least-once stream processing, schema-per-tenant isolation, and the transactional edges where those meet. I reach for Postgres primitives before I reach for another service.
- SKIP LOCKEDlease-based frontier, no external queue
- 4-stagededuplication pipeline that short-circuits
- Schema/tenantisolation with custom domains
- Python
- Django
- DRF
- FastAPI
- PostgreSQL
- Redis Streams
- Celery
- RabbitMQ
- pgvector
- WebSocket
- 01
Distributed work coordination
A crash-safe URL frontier on PostgreSQL using FOR UPDATE SKIP LOCKED - priority scheduling, exponential backoff, lease expiry and reaping, so a worker dying mid-fetch costs one lease instead of the queue.
- 02
At-least-once messaging that actually holds
Redis Streams with consumer groups, delayed retries, dead-letter queues and stale-message reclaim across cleaning → LLM → embedding stages. Idempotent handlers, because at-least-once means duplicates are a certainty.
- 03
Multi-tenant data isolation
Schema-per-tenant Postgres with custom domains, plus branch-level scoping inside each tenant and RBAC that resolves against plan, role and resource together.
- 04
Integration surface
SSLCommerz, Stripe, Twilio, Google OAuth, AWS SES webhooks and ZKTeco biometric hardware - each behind an interface so the vendor stays swappable.
LLMs are unreliable subsystems. I engineer around that.
My NLP background is academic - a Bengali↔English word alignment thesis on multilingual BERT - but my LLM work is operational: schema-validated extraction, provider fallback, cost tracking, versioned embeddings and semantic search that has to answer in production latency.
- BGE-M3versioned embeddings + pgvector search
- Bangla + ENmultilingual extraction pipeline
- Geminiagentic contact management
- LangChain
- LangGraph
- CrewAI
- Gemini API
- Ollama
- BGE-M3
- pgvector
- Scikit-learn
- Pandas
- NumPy
- 01
Schema-validated extraction
Every model call returns against a declared schema. Invalid output is a retry with repair context, not a 500 - and never an unchecked write to the database.
- 02
Provider fallback & cost accounting
Multi-provider routing with automatic failover and per-call token/cost tracking, so a rate limit degrades quality instead of taking the pipeline down.
- 03
Semantic search on pgvector
Versioned BGE-M3 embeddings served through pgvector, with SimHash banding in front so the expensive semantic comparison only runs on genuine near-duplicates.
- 04
Agentic interfaces with guardrails
Gemini-powered contact management driven by natural language and voice, where the agent proposes structured operations and the application layer validates and executes them.
- 05
Multilingual by design
Bangla and English through the same pipeline - tokenisation, dedup thresholds and embeddings all chosen to survive script and morphology differences.
If I built it, I can deploy it, watch it and roll it back.
I own the path from a merged commit to a running container behind Traefik and Cloudflare - multi-tenant provisioning included. Zero-downtime is a deployment property I design for, not a thing I hope for.
- Zero-downtimedeploys behind Traefik + Nginx
- Multi-tenantCI/CD provisioning pipelines
- Prometheusmetrics on every pipeline stage
- Docker
- GitHub Actions
- Traefik
- Nginx
- Cloudflare
- AWS
- Redis
- PgBouncer
- Prometheus
- Grafana
- Linux
- Bash
- 01
Multi-tenant CI/CD
GitHub Actions pipelines that provision, migrate and release per tenant across VPS, AWS and other cloud targets - so onboarding a customer is a pipeline run, not an afternoon.
- 02
Zero-downtime releases
Rolling container swaps behind Traefik and Nginx with health gating, so a bad build never becomes a bad request.
- 03
Capacity engineering
PgBouncer for connection pressure, Redis for read amplification and hot state, and query-level work before hardware - 10k+ DAU held without over-provisioning.
- 04
Observability
Prometheus and Grafana wired to the queue depths and stage latencies that actually predict incidents.
- 05
Release automation for packages
Tag-driven semantic versioning with generated changelogs, publishing to PyPI, GitHub Packages and Docker Hub from one workflow.
The part that outlives the sprint.
Reusable abstractions, written-down architecture, tests where they earn their keep, and mentoring so decisions survive me. A Master's in IT alongside full-time delivery, and an open-source package on PyPI that forced me to take versioning seriously.
- PyPIpublished open-source package
- Mentoringjunior engineers across two companies
- Master'sin IT, University of Dhaka
- Python
- TypeScript
- C++
- Git
- TDD
- Agile
- SDLC
- Poetry
- Jira
- GitHub Projects
- 01
Abstractions that get reused
Shared components and common layers extracted across projects - the reason feature three in a codebase costs less than feature one.
- 02
Technical leadership
Leading junior engineers, reviewing designs before code, and running knowledge-sharing so the architecture lives in more than one head.
- 03
Documentation as infrastructure
Architecture docs and decision records maintained alongside the code - onboarding time is a measurable engineering output.
- 04
Open source discipline
Data Preprocessors on PyPI: semantic versioning, automated changelogs, reproducible Poetry builds, published to three registries.
02 / Selected work
Systems, not screenshots
Platforms where the interesting part was never the UI. Each case study is written the way I would explain it in a design review.
FitsSort
Multi-tenant, multi-branch SaaS for health and fitness businesses - web, mobile and turnstile.
- 10k+daily active users
- 1 schemaper tenant
- 3clients on one API
- Python
- Django
- DRF
- PostgreSQL
- React
- +9
Read the case studyLiveTenant demoBackendFrontendMobile
03 / Track record
Where the experience came from
Four years climbing the same ladder from a different rung each time - intern, associate, engineer, senior.
Currently
Senior Full Stack Software Engineer
TechForing Ltd.Jan 2026 - Present
Technical leadership on a multi-tenant SaaS platform: breaking a monolith into services, automating tenant provisioning, and bringing AI into the internal development loop.
- Provided technical leadership to junior engineers and drove delivery of scalable SaaS platform capabilities.
- Migrated legacy monoliths into modular microservices, improving reusability, scalability and maintainability.
- Integrated AI-driven workflows into internal systems, accelerating development and improving team productivity.
Full Stack Software Engineer
Jan 2025 - Dec 2025TechForing Ltd.
Owned the communication and attendance surfaces of the product line, plus the third-party integrations everything else depended on.
Software Engineer
May 2024 - Dec 2024Nimusoft Technologies Ltd.
Moved from executing tickets to shaping systems - requirements with leadership, architecture decisions, and end-to-end ownership across several products.
Associate Software Engineer
Jan 2024 - Apr 2024Nimusoft Technologies Ltd.
Django application work with a deliberate bias toward shared, reusable building blocks.
Software Engineer Intern
Jul 2023 - Dec 2023NSU StartUps Next
Built the incubator's own platform while advising the startups inside it - a fast lesson in shipping against requirements that change weekly.
04 / Writing
Notes from inside the systems
Long-form engineering write-ups on the decisions that actually mattered - concurrency, isolation, delivery guarantees and where LLMs belong.
Selling With the Wi-Fi Down: An Offline-First POS in Tauri 2.0
A till that stops working when the internet does is a till that loses money. Here is the outbox-plus-idempotency contract that lets a Tauri desktop POS sell offline and replay safely into a Django backend - without double-charging anyone.
ReadDecomposing a Monolith Without a Rewrite
Seams before services. How to extract modular services from a legacy monolith incrementally - with the data ownership rule, the strangler pattern, and an honest list of what you give up the moment a function call becomes a network call.
ReadLeases Over Locks: A Crash-Safe Work Queue in PostgreSQL
How FOR UPDATE SKIP LOCKED plus expiring leases gives you a distributed, crash-safe work queue without adding a broker to your architecture - with the reaper, backoff and priority scheduling that make it production-grade.
ReadOpen to work
If you are building something that has to stay up, let's talk.
Senior backend, platform, or AI engineering - full time or contract. I reply to everything, usually within a day.
- Based in
- Badda, Dhaka, Bangladesh
- Timezone
- Asia/Dhaka (UTC+6)
- Working with
- Distributed teams across time zones
- Languages
- English, Bangla