Engineer · Architect · Educator
I ship software for clients in nine countries and teach 2,300+ engineers what shipped systems actually do.
Trusted across
Md Tausif Hossain
Senior Engineer · Lead Instructor
Three concurrent roles. One practice. Numbers pulled live from GitHub.
Current roles
ConcurrentTechnicalBind
Founder & Self-Employed
DevTechGuru
Tech Lead
Ostad
Lead Instructor
engineers mentored
Years Operating
Products Shipped
Engineers Mentored
About
I work across three places. At DevTechGuru I have been Tech Lead since late 2021, contributing to architecture, code review, and engineering practice on client projects across nine countries. TechnicalBind is the independent studio I started in February 2017, still operating with a small distributed team. At Ostad I teach advanced full-stack engineering. Across the three, I have mentored 2,300+ engineers, mostly on the same Python/Django and TypeScript/Next.js stack that runs my own production work. Most of what I share publicly is grounded in what I have shipped, not in what looks good in a deck.
Md Tausif Hossain
Senior Engineer · Lead Instructor
TechnicalBind, since 2017.
DevTechGuru Tech Lead, since 2021.
Ostad instructor, since 2025.
Three homes, one practice.
Years
Products
Engineers Mentored
What that looks like
Numbers from production, not demo decks. Three projects below, each still running.
AI assistant for joint-replacement planning. The model suggests, the surgeon decides. Adopted on the first ward it shipped to.
Camera-driven parking system across multiple lots. Postgres row-level locks removed the race condition that used to sell the same spot twice.
University publishing platform serving 2M+ reads a day. Editors publish in fifteen minutes instead of two days, and engineering is out of the publishing loop.
Work & Impact
A handful of projects I led from kickoff to launch. Tap any card for the full story.
Showing 6 of 10
04more · click to reveal
Quick reality check
Send 3 lines. I reply with whether I am the right pair of hands.
Blog
Operating rules
Four rules. Each earned on a project that did not behave the way the diagram said it would.
On the smart-parking system, the same spot was being sold twice under load. A row-level lock on the reservation table removed the race in one migration. No Kafka, no service split, no rewrite. A single Postgres instance has handled every project I have shipped. Split it only when a query you cannot rewrite is the bottleneck.
OrderBind ships every table behind a row-level security policy keyed on tenant id. Application code that forgets the filter still cannot leak data across tenants. RLS is slower to author than an `if` statement in a handler, and it is the only check that survives a refactor, a new endpoint, or a tired Friday merge.
In the AI orthopaedics tool, every suggestion is visible, editable, and overridable in one click. Nothing proceeds without a surgeon’s sign-off. The model earns trust by being legible, not by being right; an opaque AI that is correct 95% of the time is unusable in a clinical setting, but a transparent one at 80% gets adopted on the ward.
Before reaching for a queue, a cache, or a new service, try the primitive the database already gives you. Row locks before mutexes. Indexes before read replicas. Materialised views before cron jobs. Most production bugs I have fixed were a feature in Postgres I had not used yet.
Experience
An independent studio, a tech-lead seat, and a teaching role. All active today.
Feb 2017 - Present (9+ years)
Independent Studio
I started TechnicalBind in February 2017 as an independent software studio. Over time it grew from one-person work into a small distributed team building production web and mobile platforms. Examples include a CMS serving 2M+ daily reads and fintech apps published to Google Play. Mentoring junior contributors on the team has been part of the work from year one.
Result
Nine years of continuous operation. Runs in parallel with my Tech Lead role at DevTechGuru since 2021.
Nov 2021 - Present (4.5+ years, concurrent with TechnicalBind)
Corporate Track
I joined DevTechGuru full-time in late 2021 as Tech Lead while continuing to run TechnicalBind in parallel. The team ships software for clients across nine countries, covering healthcare AI, university CMS, fintech, smart parking, ERP, and adjacent work. Day-to-day I focus on architecture review, code review, and helping the engineers around me grow into harder problems.
Result
Contributed to the move toward trunk-based development, GitHub Actions CI/CD, RFC-first design, and async written-communication standards. The team operates async across nine timezones.
June 2025 - Present (concurrent, part-time)
Teaching
I teach advanced system design and full-stack engineering at Ostad, mostly to working developers preparing for their first remote role outside Bangladesh. The course material is drawn from the same systems I ship at DevTechGuru and TechnicalBind, so what students learn is closer to production than to slideware.
Result
Some students have gone on to ship to production at companies they had not previously applied to. Lessons from their work also feed back into how I review code at the day job.
Feb 2021 - May 2021 (4 months)
Systems Foundation
Short pre-graduation internship. Built a face-recognition attendance system for a wellness chain. My first production system that a business depended on.
Result
Learned where hardware and software diverge in practice in ways university coursework did not cover.
Mid-session
Where the easy answer breaks down.
Hard problems solved
The obvious approach failed, the correct fix had a cost, and the tradeoff was worth taking. Both are still in production.
Situation
Surgeons are cautious about AI in clinical settings. A black-box prediction on a hip replacement is a liability concern.
Response
I built the UI around one rule: the model proposes, the surgeon decides. Every suggestion is visible, editable, and overridable with one click. Nothing proceeds without sign-off.
Result
Adoption reached 100% on the first ward where it was deployed. Pre-operative planning time dropped by roughly 40%, and surgeons reported higher trust in the resulting plans.
Situation
During traffic spikes, the booking system was selling the same spot to two drivers. Both reservations validated against an empty slot before either write committed.
Response
Wrapped every reservation in a row-level Postgres lock and added a sweeper for stale holds. Cost: 30ms per booking. Benefit: the race condition is gone.
Result
Zero double-bookings across 100,000+ reservations since launch. The 30ms is below user-perceptible latency, and the manual reconciliation that used to follow each incident is no longer needed.
Get in touch
Bangladesh (GMT+6). I reply within a day.
Newsletter
Occasional posts on shipping software, leading engineers, and what production actually teaches.
