Portfolio — 2026 · Birmingham, UKOpen to Work

( 01 ) Identity

QAYYUM

BOKHARI

Full-Stack Developer & AI Engineer

Building production software — React, Django, and LLM integrations.

RESUME GET IN TOUCH

Scroll

( 02 ) About

About Me

Commercial Experience. Production-Ready. Always Building.

I'm a Software Engineer with commercial experience building production web applications and AI-powered systems. I've delivered measurable results including a 23% reduction in client return rates through predictive analytics and a 35% improvement in backend performance through strategic optimisation.

My journey started with cold-calling businesses on Google Maps and offering to build their websites. That taught me that shipping matters more than perfection. Since then I've built production inventory systems with Django and Laravel, REST APIs with JWT auth and Redis caching, and deployed AI agents and RAG chatbots using OpenAI and Hugging Face.

Currently studying BSc AI & Robotics at Aston University (graduating 2027). Full right to work in the UK. Actively seeking Software Engineer, AI/ML Engineer, or Full Stack Developer roles.

Years Building

20+

Projects Shipped

Right to Work

Download CV Get in Touch

( 03 ) Expertise

Skills & Expertise

Backend

PythonDjangoLaravelPHPNode.jsPostgreSQLMySQLMongoDBSQLite

Frontend

JavaScriptReactHTML5CSS3jQueryTailwind CSS

AI / ML

Hugging Face TransformersScikit-learnFAISSOpenAI APILarge Language Models

DevOps & Cloud

DockerGitCI/CDAWSVercel

Testing & Practices

Unit TestingIntegration TestingTDDAgileScrumREST API Design

Loading education…

Loading certifications…

( 04 ) Work

Featured Projects

Real problems. Real solutions. Real results.

Portfolio Optimisation Platform

Hybrid portfolio optimisation platform integrating classical financial models (CAPM, Black-Litterman, Monte Carlo simulation, Value-at-Risk) with machine learning (Linear Regression, Random Forest, Gradient Boosting). ML forecasts are fed as probabilistic "views" into the Black-Litterman framework, producing portfolios that consistently outperform both classical and ML-only approaches. Features an interactive dashboard with efficient frontier visualisations, allocation dynamics, performance metrics, and stress testing under bearish, bullish, and crisis market scenarios.

Problem

Classical mean-variance optimisation relies on assumptions like normal returns and stable covariances, producing fragile and concentrated portfolios. Pure ML approaches improve predictions but generate unstable, high-turnover allocations that are impractical to deploy.

Result

Hybrid framework achieved a 1.20 Sharpe ratio (vs 0.85 classical, 1.05 ML-only), Sortino ratio of 1.55, lowest Value-at-Risk at -1.8% daily, and 25% lower drawdowns under crisis stress tests — the best balance of returns, stability, and risk control across all metrics.

PythonScikit-learnLinear RegressionRandom ForestGradient BoostingBlack-LittermanMonte CarloCAPMVaRCVaRPandasNumPyMatplotlib

Paper

AI-Powered CV Chatbot

Intelligent chatbot using Retrieval-Augmented Generation (RAG) architecture to provide context-aware responses about CV content. Integrated OpenAI GPT-3.5 API and Llama 2 via Hugging Face Transformers.

Problem

Standard LLMs hallucinate when answering questions about specific CV details — confident, wrong answers.

Result

92% response accuracy using RAG + FAISS vector search to ground every answer in real CV data.

PythonDjangoOpenAI APIHugging FaceFAISSAWS EC2Docker

Code

Royal Tap E-Commerce

Custom e-commerce solution built for a UK business with payment gateway integration, full mobile responsiveness, and SEO optimisation. Achieved a 90+ Google PageSpeed score.

Problem

The client needed a performant online store that could handle payments securely without relying on expensive SaaS platforms.

Result

Delivered an end-to-end e-commerce site with integrated payments, 90+ PageSpeed score, and full SEO configuration — handling real transactions in production.

WordPressCustom PHPWooCommerceSEO

Code Live Demo

Shuttercraft Boats

Full-stack business website with a content management system and customer inquiry workflow. Includes RESTful API endpoints, database schema design, and server-side form validation with automated email notifications.

Problem

The business had no online presence and was relying entirely on word-of-mouth for customer inquiries.

Result

Built a professional site with a CMS for self-service content updates and an automated inquiry system that routes customer requests directly via email.

LaravelMySQLREST APIPHP

Code Live Demo

Retro App

Secure team collaboration platform for cross-functional teams. Designed with a clean REST API, JWT auth, and a Redis caching layer for high-traffic scenarios.

Problem

Existing collaboration tools were sluggish under load — 2+ second response times were killing team flow.

Result

60% faster response times (2.1s → 0.8s) after optimising queries and implementing Redis caching.

DjangoPostgreSQLRedisJWTREST APIDocker

Code

Interactive Chemistry Buffer Simulator

A React-based simulation tool for chemistry education. Features live Henderson-Hasselbalch pH calculations, interactive buffer component controls, dynamic titration curves, a visual pH scale with real-time indicator, and buffer effectiveness status monitoring.

ReactRechartsVite

Code Live Demo

FotoMarketplace

A photography and videography marketplace connecting creators with clients. Features photographer profiles, portfolio galleries, booking system, and search with filters.

ReactTypeScriptViteTailwind CSSSupabase

Code

Cohere RAG Chatbot

Document Q&A chatbot powered entirely by Cohere's AI stack. Upload any PDF or text file and ask questions — every answer is grounded in the document with inline citations showing exactly which passages the model used. Built without LangChain or any framework — raw pipeline from scratch.

Problem

LLMs hallucinate when answering document-specific questions because they rely on parametric memory rather than the actual content.

Result

5-step RAG pipeline (embed → FAISS → rerank → generate) with character-level inline citations, achieving precise retrieval via cross-encoder reranking over raw vector similarity.

CohereRAGFAISSNext.jsTypeScriptFastAPIPython

Code Live Demo

MapCanvas

Interactive e-commerce platform that lets users design and buy custom maps of their favourite village, town, or city. Built with MapBox for map rendering and Gelato API for print-on-demand fulfilment. Users customise their map style, framing, and location then order physical prints delivered worldwide.

Result

Generated over 40,000 maps across 35+ countries since launch.

ReactMapBox APIGelato APIE-commerceVercel

Code Live Demo

( 07 ) Writing