I am the engineering department at NV Store, a Bangladeshi e-commerce business selling home decor and appliances. I built and run the storefront, the ERP and CRM, CI/CD, monitoring, backups, databases and the phone system.
Every parcel below is something I built and still run in production.
The customer-facing e-commerce site, in production at nvstore.shop.
I built the full site: catalog, product pages, cart, checkout and the admin side the team uses every day. It shares its data with the in-house ERP, so stock and orders stay in one place.
Four stores, each picked for one kind of job.
Three features in daily use. Each one removes a manual step, and a person stays in the loop where mistakes would cost money.
Turns raw product input into a polished description. Already used on nvstore.shop.
Staff enter what they know: name, material, size, color, a few notes. The system builds a prompt from a fixed template, generates a draft, and puts it in front of a person to accept or edit before it goes live.
product notes
name: Ceramic table vase material: ceramic, matte glaze color: white size: 30 cm notes: gift box, good for living room
generated description
A matte white ceramic vase, 30 cm tall, that fits a living-room shelf or a dining table. The soft glaze works with fresh or dried flowers. Arrives in a gift box, ready to give.
Moderators paste a messy customer message and the system fills the ERP order fields.
The moderator pastes the raw message into the order screen. A model extracts name, phone, address and order details into a fixed schema. Validation checks the phone format and matches the address against known districts and areas. The form is filled, and the moderator reviews it before saving.
pasted message
vai amar nam Rahim Example 01700-000000 basa house 00 road 0 example para mirpur 10 dhaka 2 ta lagbe, white ta
filled order fields
name Rahim Example phone 01700-000000 address House 00, Road 0, Example Para area Mirpur district Dhaka quantity 2 note white
Text search, image-to-image and image-to-product search with an embedding model and MariaDB as the vector database.
Each product’s images and text are passed through an embedding model and stored as vectors in MariaDB next to the product row. A search query, typed or uploaded as a photo, is embedded the same way, and the nearest products come back ranked by similarity.
search
query: "white round vase" (text) or: photo of a vase (image)
results
1 Ceramic table vase, white 2 Round glass vase, frosted 3 Ceramic bud vase, ivory ranked by embedding similarity
In-house system for orders, inventory, customers, roles and permissions.
I designed the data model and built the application around how the team actually works: a moderator takes an order, the warehouse sees it, stock updates, and the customer record keeps the history. Roles and permissions decide who can view, edit or approve each step.
Replies to comments on NV Store's Facebook page and moves price or detail requests to the inbox.
An n8n workflow listens for new comments on the page. It checks whether the comment asks about price or details. If so, it replies publicly and sends the person a private message so the conversation continues in the inbox.
FusionPBX on a VPS for NV Store's customer calls, with IVR and queues.
I set up and run FusionPBX on a VPS for NV Store’s support line: IVR menus route callers, and queues hold calls until an agent is free. Running it in production led to my upstream fix for the FIFO queue (FusionPBX pull request #8225).
CI/CD, monitoring, encrypted automated off-site database backups and database administration.
Every push runs through GitHub Actions, builds a container image, and deploys it. Monitoring checks the public services and alerts me when something breaks; the 90-day history is public on the status page. Databases are backed up automatically, encrypted, and copied off-site.
Public repositories, plus one upstream fix.
Production code belongs to NV Store and stays private, and so does customer data. So this page shows the problem, the approach and the outcome instead of source code. The live store is the demo. Code I can share is in the open-source toolbox above.