Results · Recent work

Systems quietly earning their keep.

Client projects, internal tools, and hackathon builds — all designed to solve one problem properly.

15+ hrs
Saved / week — hiring
0
Manual data entry
95%+
ML accuracy
100k+
Orders analyzed
● Architecture Client project Hiring automation

QALAM Architects went from 15 hours of resume screening a week to under 3.

15+ hrs
Saved per week
14 days
From kickoff to live
−80%
Screening time

The problem

Every week, QALAM's partners spent 15+ hours personally reading applications, scheduling interviews with people who turned out to be unqualified, and playing admin instead of doing architecture.

What we built

A pre-screening pipeline that parses every applicant's resume and portfolio against weighted criteria — experience, software, location, visa. Only shortlisted candidates surface in the partners' inbox. Everyone else gets a professional auto-response.

The result

They now interview fewer candidates but hire better — and got their Saturdays back. The system runs without their attention; they just get a weekly shortlist.

“It pays for itself the first month. After that it just quietly works.”
Amirul F. · Managing Partner
● Document ops Internal tool OCR + AI

From stacks of PDFs to a clean spreadsheet — seconds per document, zero keystrokes.

0
Manual data entry
< 10s
Per document
Auto
Archive & filing

The problem

Most businesses get a weekly flood of invoices, receipts, forms, purchase orders. Someone has to open each, copy data into a sheet, rename the file, archive it. It's slow and error-prone.

What we built

A watcher on an input folder. Documents are parsed (OCR + LLM), the fields that matter are extracted, a row is appended to Google Sheets with correct typing, and the source file is renamed and moved to an archive folder — searchable forever.

The result

What was half a day's work every week is now invisible. Drop a file in; look at the sheet. Anomalies get flagged for a 3-second human check.

● E-commerce / AI Hackathon build XGBoost · recsys

A live analytics dashboard that predicts cart abandonment before the customer leaves.

0.95+
ROC-AUC accuracy
100k+
Real orders analyzed
3-in-1
Tools unified

The problem

Stores lose revenue to cart abandonment every day — usually they find out after the sale's already gone. By then it's too late to win it back.

What we built

A dashboard trained on 100,000+ real orders that flags carts likely to abandon in real time, recommends products tuned to the individual, and gives operators a revenue view with purchase heatmaps and regional breakdowns.

The result

A single tool an operator can use to do something about abandonment in-session — not just report it after the fact. Built in a weekend.

Your industry is probably on our list

Want results like these?

Every project starts with a free walkthrough. Tell us the problem; we'll tell you honestly if we can fix it.

Book a free walkthrough