I build the backend behind AI products.

Case study·2025–26Featured

Auctoritas

Document intelligence and staff training platform, backend built solo

A multi-tenant SaaS where organisations upload agreements and policies, ask questions with page-level citations, compare documents, and train staff through video playlists with AI-generated quizzes. I designed and built the entire backend, from the ingestion pipeline to billing and deployment.

• Async PDF ingestion with Celery and Redis: heading-based chunking, live progress over SSE, retries and a dead-letter queue
• Streaming RAG chat over Weaviate with citations, token-budgeted context and a prompt-injection guard
• Stripe subscriptions with plan limits, per-organisation overrides and multi-tenant roles across organisations and worksites

Built with

• FastAPI
• MongoDB
• Celery
• Redis
• Weaviate
• LangChain
• OpenAI
• Stripe
• AWS S3
• Docker

Private repository · architecture walkthrough on request

Featured

Final-year project · Team of 3·2025–26

VECTA

Voice-enabled crypto trading assistant

A web app and Chrome extension that give crypto price forecasts and buy, sell or hold signals across 1-hour, 4-hour and 24-hour timeframes, with voice commands. I led the project and built the ML service: a Temporal Fusion Transformer with quantile loss that forecasts price ranges instead of single points, a signal engine on top of it, and the FastAPI service the main backend calls for predictions. The research paper is under review.

• Feature pipeline over Binance OHLCV data: RSI, MACD, Bollinger Bands, with separate scaling per feature group
• Temporal Fusion Transformer (PyTorch Forecasting) predicting quantile ranges over a multi-step horizon
• FastAPI ML service consumed by a NestJS backend, alongside Whisper and LLM-based voice intent handling

Built with

• PyTorch Forecasting
• Temporal Fusion Transformer
• FastAPI
• NestJS
• React
• PostgreSQL
• Redis

Featured

Product·2025

Pictofy AI

AI pet portraits, live on both app stores

An AI image-generation PWA published to the App Store and Google Play, with Stripe and RevenueCat handling subscriptions and one-off purchases.

Built with

• Next.js
• PWA
• Node.js
• Stripe
• RevenueCat
• AWS S3

Visit site

Client work·2025

Phone Fixer POS

Point-of-sale system in daily use at a phone repair shop

A Next.js point-of-sale system for a phone repair business, covering inventory, categories, sales and a dashboard with revenue and low-stock alerts.

Built with

• Next.js
• TypeScript
• PostgreSQL
• Vercel

Client work·2024

Impacting Life Productions

Media platform for a UK film production company

A MERN media platform with a custom CMS for the production team, dynamic routing for film and series pages, and a media library backed by S3.

Built with

• React
• Node.js
• Express
• MongoDB
• AWS S3

Visit site

Earlier deep learning work: melanoma classification (Kaggle notebook), image captioning and age & gender prediction. More on GitHub. Most of my production work is under NDA, and I’m happy to walk through the architecture on a call.

1. Lead AI-Product Engineer

   Cognara Training

   B2B AI training and document-intelligence platforms

   Jul 2025 – PresentLahore, Pakistan

   • Team leadership — Lead a team of three junior developers — running code reviews, guiding architectural decisions, and enforcing coding standards across the codebase.

   • Retrieval architecture — Built RAG pipelines with LangChain and Weaviate, using Unstructured hi-res parsing and heading-based chunking for scanned and complex PDFs.

   • Real-time document features — Shipped an interactive agreements chat and a live side-by-side document comparison, streamed over SSE.

   • AI-generated assessments — Designed a pipeline that generates quizzes from training-video transcripts, with a manual override so instructors keep editorial control.

   • Backend & infrastructure — Architected async task processing with Celery and Redis across MongoDB and Weaviate, deployed with Docker Compose and GitHub Actions on AWS.

   • Frontend consolidation — Merged separate onboarding and dashboard frontends into one React codebase with a shared design system.

   Technologies used

   • Python
   • FastAPI
   • Weaviate
   • MongoDB
   • Celery
   • Redis
   • React
   • Docker
   • AWS

2. Independent Full-Stack Developer

   Freelance & personal products

   Client platforms and self-published apps

   2024 – PresentRemote

   • Pictofy AI — Built and published an AI pet-portrait PWA to both the App Store and Google Play, with Stripe and RevenueCat billing.

   • Impacting Life Productions — Delivered a MERN media platform for a UK film production company, covering custom backend architecture, dynamic routing, and content data models.

   • PhoneFixer POS — Built a Next.js point-of-sale system in production use at a repair business, handling inventory, ticketing, and sales reporting.

   Technologies used

   • Next.js
   • Node.js
   • TypeScript
   • MongoDB
   • Stripe
   • RevenueCat
   • PWA

3. BS Computer Science

   Superior University

   Graduated 2026Lahore, Pakistan

   • Final-year project — VECTA — Led a three-person team building a voice-enabled crypto trading assistant. Built the ML service: a Temporal Fusion Transformer with quantile forecasting, signal generation, and a FastAPI service. Research paper submitted and under review.

   Technologies used

   • PyTorch
   • Temporal Fusion Transformer
   • NestJS
   • FastAPI
   • Research

I take on freelance projects and I’m open to full-time roles, remote or relocation. If you’re building an AI product or need a backend that holds up in production, email is the fastest way to reach me.

Location

Lahore, Pakistan

Timezone

PKT (UTC+5) · overlaps EU & partial US

Looking for

Freelance projects · full-time roles