·1 min read
What Stat 2 at BUP actually taught me about building real analytics tools
Business Analytics (Stat 2) at BUP pushed me to turn a classroom survey into a real Supabase-backed web app in R — here's what stuck.
Learning
Blog
Loading...
Ahsan Habib Akik — forward deployed engineer
Embed — Prototype — Ship — Measure.
I drop into the messy problem, ship production code in days, and prove the outcome in numbers. Not a freelancer — an engineer who owns delivery.
// how I work
Embed and prototype feed a ship hub; shipping feeds measurement; the outcome loops back into the next embed. The engagement is a cycle, not a hand-off.
// the proof, not the pitch
Cart, Stripe payments, inventory, and an operator dashboard — Bengali-localized, shipped to prod.
n8n + Claude + Gemini wired into real pipelines — Cal→ClickUp, content gen, daily briefs.
Three services on one Dokploy droplet — DigitalOcean + Cloudflare + Docker. I own uptime.
Bespoke design tokens, generated visuals, and a library of custom animated UI components.
AI Buddy build// deployments — live in production
Real sites in production right now. Open any of them — the build is the proof.
Video production studio
Cinematic brand site for a video studio — GSAP motion throughout, a custom admin CMS to manage page videos, and an editorial “our process” flow.
SaaS marketing site · Canada
Marketing site for dojo-management software — feature storytelling, a Ghost-powered blog, and wired-up email + newsletter automation.
Personal brand · AI strategist
Personal site for Sohan Haidear — AI strategist & consultant. Services, tiered packages, GSAP motion, and a Ghost-backed blog.
operate
Not a demo — a self-hosted n8n workflow on my own Dokploy droplet, wiring Claude + Gemini into a daily-brief agent. It embeds with the data, drafts, ships to ClickUp + email, logs every run, and evals its own output.
embed→prototype→ship→operate→measure
Design sense is part of the job — an FDE ships things people want to use. Pick a brand identity I've built to see its real palette paint the stage, or switch to Generate yours and roll a fresh, harmonious five-role system. Every pixel of the art is computed from these colors in your browser.
jewelry e-comm · pearl-luxe identity
The real Rupomoti identity — click any swatch to copy its hex. Want to roll your own? Hit .
// field notes
Writing on automation, AI agents, self-hosted infra, and the messy reality of shipping software solo.
All postsHow I ship: automation, AI agents, and self-hosted infra. No spam — opt out any time.
// faq
The things people usually ask before we start. Anything else — just send a note.
A front-line engineer who embeds directly in your problem — your stack, your data, your users — and owns delivery end to end: maps the problem, prototypes, ships to production, operates it, and proves the business outcome. Less "hand off a ticket," more "sit with the team and make it work."
A working prototype usually lands in days, not weeks. Most engagements start within 2–3 days of a scoping call, and I ship to production continuously rather than saving everything for one big launch.
Production software end to end: full-stack web and e-commerce (Next.js, React, Node, Postgres/Mongo, Stripe), AI automation and agents (n8n + Claude + Gemini), and the infrastructure to run it — often self-hosted on Docker/Dokploy with Cloudflare.
Yes. I co-founded an e-commerce startup and ship for founders and small teams who need an engineer who can also talk to customers and reason about the business — not just close tickets.
Scoped to the outcome and the work involved — transparent and value-based, agreed up front on a free 30-minute call. No hidden costs, no surprise invoices.
Based in Dhaka, Bangladesh, working remotely with teams worldwide. Async-friendly, with overlap hours arranged per engagement.
embed → prototype → ship → operate → measure
I embed with the messy reality, prototype fast, and ship production code that moves a real number. Book a 30-minute call to scope it, or send the details and I'll come back with a plan.
I sit with the problem — your stack, your constraints, your users.
A working build in days, not a deck. We iterate on the real thing.
To production, then the number that proves it worked.