Skip to content
<ali />

/*software engineer*/

Building backend, data and AI systems that hold up in production.

I design and build APIs, data pipelines and evaluation tooling with Python and PostgreSQL, then use TypeScript when the product needs an interface.

Explore selected workRead engineering notes
Python
FastAPI
PostgreSQL
Docker
Prefect
LLM APIs
TypeScript

/*selected work*/

Systems with enough depth to inspect.

A data platform, an LLM evaluation framework, and production client work that show different parts of how I build.

All projects
End-to-end Retail Ops platform architecture showing CSV ingestion, Prefect orchestration, Bronze/Silver/Gold transformations, validation and quarantine handling, analytics generation, and FastAPI serving endpoints.

data engineering / backend

01

Retail & Operations Intelligence Platform

Python ELT platform with PostgreSQL, Prefect orchestration, medallion architecture, and a FastAPI analytics REST API. Containerised with Docker Compose.

PythonPostgreSQLPrefectFastAPI
case study
Comparative LLM evaluation framework measuring GPT-4o and Gemini 2.5 Flash across financial sentiment classification accuracy, macro F1, per-class F1, latency, confusion matrices, and hallucination-aware Q&A performance.

machine learning / evaluation

02

LLM Benchmarking Framework for Financial Sentiment Analysis

BSc final-year project benchmarking GPT-4o vs Gemini 2.5 Flash on financial sentiment datasets with F1, latency, hallucination analysis, and McNemar significance testing.

PythonOpenAI APIGemini APIPandas
case study

client work / frontend / production

03

Sana Wajid Photography

Production photography portfolio for a Manchester-based wedding photographer, built with Next.js to match a Pixieset brand identity with improved responsiveness, SEO, and contact reliability.

Next.jsTypeScriptTailwind CSSFramer Motion
case study

/*latest note*/

Writing down what I learn while building.

5 min read

Benchmarking GPT-4o and Gemini 2.5 Flash for Financial Sentiment

Notes from building a reproducible LLM evaluation pipeline across Financial PhraseBank and FiQA, including classification quality, latency, hallucination checks, and McNemar testing.

#llm-evaluation#python#nlp