Menu

About

I'm a software engineer based in Lahore, Pakistan. I studied computer science at Georgia Tech, where I spent nearly three years as an undergraduate researcher in computer vision and urban mobility. My main project was detecting crosswalks in noisy, low resolution aerial imagery: by rebuilding the training dataset and refining the pipeline, I took detection accuracy from a 25% baseline to 60%. I also worked on accessible navigation and on modelling the cost of pedestrian infrastructure, which meant collecting field data across several cities in person.

Before that I interned at NETSOL Technologies, where I built and deployed an LLM based resume screening service.

Since graduating I've moved back to Lahore and started building. I co-founded LEPUS, an Amazon business selling in the US and UK, and wrote LEPUS OS, the accounting and inventory system it runs on. I'm also building Car Passport, a vehicle history archive for Pakistan's used car market.

I like owning the whole thing: the data model, the edge cases nobody else wants, and the last pixel.

Resume (PDF)

Experience

  1. to present

    Co-founder, Engineering and AccountsLEPUS

    Lahore, Pakistan (remote)

    An Amazon FBA business with US and UK entities.

    • Built LEPUS OS, the accounting and inventory system behind the business (TypeScript, Node.js, Next.js, React, PostgreSQL), with Amazon SP-API synchronisation that resumes from checkpoints instead of duplicating records.
    • Implemented batch level FIFO landed costing: freight and import duty are allocated to individual stock batches, and inventory is held as an asset until sale, so every SKU's margin includes product, freight and duty cost.
    • Built settlement reconciliation that splits each payout into sales, refunds, fees, adjustments and reserves, and compares it with what the recorded events say it should be.
    • Recorded refunds as financial events and physical returns separately, only with evidence, so a refunded unit never goes back into stock by assumption.
    • Amounts in other currencies are converted using only the exchange rate actually applied, never an estimate.
    • LEPUS OS has 82 passing automated tests.

  2. to

    Undergraduate Researcher, Computer Vision and Urban MobilityGeorgia Institute of Technology

    Atlanta, Georgia

    • Raised crosswalk detection accuracy on noisy, low resolution aerial imagery from a 25% baseline to 60% by rebuilding the training dataset and refining the ML pipeline.
    • Worked across three research threads (crosswalk detection, accessible navigation, pedestrian infrastructure cost modelling), carried out multi-city field data collection, and built tools estimating equipment, labour and travel costs.
    • Built reproducible reporting and visualisation workflows used by a multidisciplinary research team.

  3. to

    AI/ML Intern, NETSOL Technologies (NASDAQ: NTWK), Lahore.

    Built and deployed an LLM based resume screening service.

Education

  1. BS Computer ScienceGeorgia Institute of Technology

    Atlanta, Georgia

    Coursework: Data Structures and Algorithms, Machine Learning, Computer Vision, Computer Graphics, Database Systems, Human Computer Interaction, Cognitive Science.