Reyan Tariq · A story in six chapters
An unoptimised process. A noisy dataset. A decision waiting on numbers nobody trusts.
Reyan Tariq
I'm a software engineer at Standard Chartered in Karachi. I find the slow, manual, copy-paste parts of how a business runs, and I build the thing that makes them disappear.
I · Origin
Karachi · December 2020
Before university, I started ZNR, a small business selling tech accessories. In one quarter of 2021 it fulfilled 100+ orders and brought in PKR 240,000.
Every order was a tiny operation: stock, payment, delivery, a customer waiting. I learned early that operations isn't a department. It's a promise you keep, one order at a time.
100+
orders in one quarter
PKR 240K
revenue, Q2 2021
2 yrs
running it
II · Build
SZABIST · Computer Science · 2021 – 2025
B.Sc. Computer Science: 3.7/4.0 CGPA, Dean's List, merit scholarship. The grades were the receipt. The education was shipping things people could actually use.
Client project · Live
Commerce for a mentoring practice: a shop for physical and digital products, courses, paid one-to-one consultations and Stripe checkout, running on a Medusa commerce backend.
Final-year project
An events app that matches organisers with the right vendors. A Python recommendation engine does the matchmaking; vendors get a dashboard for bookings, revenue and ratings.
Student Work Experience Engagement Platform · Live
Bridges the gap between students and employers: real work from real companies, rewarded through a token system students can show off.
83.49%
accuracy
NLP · Financial services
Sorts consumer financial complaints (CFPB data) into product categories: credit cards, mortgages, debt collection, student loans. Five models compared; a tuned SVM won.
F1 score by class
Weakest on the smallest classes. The data was imbalanced, and the README says so.
NLP · Transformers
Fine-tuned German BERT on 2,429 parliamentary speeches to predict a speaker's political orientation. 89% accuracy overall.
III · Signal
Standard Chartered · Technology Intern · Summer 2024
I built AI transcription and a Linear SVC classifier that works out what a call is about and sends it where it should go. Call routing improved by 50%.
Same summer: exploratory analysis of incident reports that cut resolution time by 15%, and the SIT/UAT environments behind the Payment-to-Merchant QR rollout, with testing throughput up ~40%.
100K+
calls a month
50%
better routing
−15%
resolution time
+40%
test throughput
NayaPay · DevOps Trainee · Winter 2024
At NayaPay I deployed and maintained Java financial applications through four environments, and watched them with Checkmk, Grafana and Prometheus so people's money kept moving.
IV · The Machine
Standard Chartered · Software Engineer · July 2025 → now
45,000 transactions every month, reconciled for Pakistan's federal tax authority. By hand, it took three hours, every single week.
A Python rule engine that consolidates the month's datasets, checks withholding-tax applicability and reconciles every row. Three hours became minutes.
A daily operations workflow went from 60–90 minutes to seconds. I extended the same approach to IFRS 9 Islamic Banking validation and technology-obsolescence reporting.
Credit card and personal loan fulfilment, around 500 applications a month, moved off legacy processes and onto one central platform.
Daily operations workflow
01:30:00
I built a self-service regulatory data portal so non-technical teams could run their own compliance queries. That ended ~10 hours a week of manual data requests.
≈0
hours a year, handed back to people
My estimate from the numbers above: ~143 h of reconciliation (3 h a week, now minutes) + ~250 h of the daily workflow (60 min × ~250 working days) + ~520 h of data requests (10 h × 52 weeks).
V · Proof
Competitions & leadership
Standard Chartered Graduate Development Event (GDE)
2026
Standard Chartered aXess Academy Bootcamp
100+ graduates across SCB markets
ZAB-E-FEST Inter-University Hackathon
Machine learning competition
Also: Assistant Director IT, SZABIST ACM (Association for Computing Machinery) · Google Data Analytics certificate · Stanford Machine Learning Specialization
VI · Next
I'd like to find it, whether it's hiding in operations, strategy or code. Let's talk.
Designed and built with Next.js, React Three Fiber and 45,000 particles.
