aligpk

Full Stack Engineer | Lahore, Pakistan

Ali
Abdullah

Sole engineer behind Mehta Brothers' entire digital infrastructure: a manufacturing ERP, a lab services SaaS, a sales CRM, and machine learning APIs, all designed, built, and run in production by one person.

Message me on LinkedIn contact@alig.pk
15production systems built solo
~225Klines of code in production
100sof external customers served
4+years of professional experience

Systems Ledger

BatchPRO

Manufacturing ERP Operational · 2024 -

The company's core ERP: bag level inventory traceability across the plant and linear programming optimization for least cost feed formulation.

Largest system in the portfolio · Next.js · Firebase · LP optimization

Case study

Context

The plant ran production planning, inventory, and feed formulation on paper registers and spreadsheets.

Problem

A quality complaint couldn't be traced back to the raw material bags that went into a batch, and formulas were costed by hand which was slow, and rarely the cheapest valid mix.

What I built

An ERP where inventory is modeled at the individual bag level: every production batch records exactly which bags it consumed, giving full forward and backward traceability. On top sits a least cost formulation engine, linear programming takes nutritional constraints and current ingredient prices and solves for the optimal mix.

Key decision

Treating the bag not the lot or the SKU as the atomic inventory unit. It made the data model heavier, but it's what makes batch level recall and shrinkage analysis possible at all.

Outcome

The company's core operational system since late 2024, and the largest codebase in my portfolio.

Insight

Lab SaaS / LIMS Operational · 2023 -

A multi app laboratory information system with credit based billing and a multi stage lab report lifecycle, from sample intake to verified results.

Serving hundreds of external customers · Next.js · Firebase · Algolia

Case study

Context

The in house lab tests feed and raw materials for paying external customers not just internal QC.

Problem

Reports moved through testing, review, and delivery by hand; billing lived in a separate ledger; customers had no way to check status or history.

What I built

A multi app SaaS: A customer portal plus internal lab applications with credit based billing (customers hold prepaid credit, each test deducts) and a multi stage report lifecycle from sample intake to verified release. Algolia powers instant search across samples and reports.

Key decision

Modeling the report lifecycle as a stage gated state machine with role based transitions, so an unverified result can never reach a customer by construction, not by policy.

Outcome

Hundreds of external customers now self serve their reports, history, and balance.

BIRD

Sales CRM Operational · 2026 -

A field sales CRM with dual authentication, drag and drop visit scheduling, and a multi stage opportunity pipeline used by the sales team daily.

Built and shipped solo, end to end · Next.js · Firebase

Case study

Context

A field sales team managing visits and deals through phone calls and personal notes.

Problem

No shared pipeline, no visit planning, and no visibility for management into what was scheduled, moving, or stuck.

What I built

A CRM with a drag and drop scheduling board for planning field visits, a multi stage opportunity pipeline, and dual authentication paths to match how different users actually sign in.

Key decision

Designing the scheduler around how the team already planned their week drag a customer onto a day instead of form based entry, which is where most field team CRMs lose adoption.

Outcome

The sales team's daily driver since early 2026, designed, built, and shipped by one person.

NIR Feed Spectrum Analysis API

Machine Learning Operational

TensorFlow/Keras regression models that predict feed composition directly from near infrared spectrometer files, integrated with Bruker OPUS hardware.

Lab instrument → ML prediction pipeline · Python · TensorFlow · Keras

Case study

Context

The lab runs a Bruker NIR spectrometer; full wet chemistry analysis of ingredients like soybean meal is slow and costly per sample.

Problem

Turning a raw spectrometer scan into a composition estimate normally requires proprietary calibration software and a specialist to run it.

What I built

A Python API that parses Bruker OPUS spectrum files, runs them through TensorFlow/Keras regression models trained on the lab's own reference results, and returns composition predictions over REST.

Key decision

Training on the company's own historical lab data instead of buying generic calibrations the models reflect the actual raw material sources the company buys from.

Outcome

Composition estimates in seconds from a single scan, with wet chemistry reserved for verification.

…and eleven more in production, An AI trade show kiosk with a multilingual RAG chatbot, a research trial platform with QR scanned data entry, an HR performance system, data scraping pipelines, and the network infrastructure underneath them.

Core stack

Next.js·TypeScript·Firebase·Python·TensorFlow·OpenAI APIs

Need one engineer who can own the whole stack?

From the database schema to the ML model to the screen the customer sees, I've built and run all of it. The fastest way to reach me:

Message me on LinkedIn contact@alig.pk