Skip to content

Work

Every public project, grouped by area. Live demos run in your browser; results are quoted from the notebooks and READMEs that recorded them.

42 projects shown

Agentic AIRAG and Knowledge

Veriflow

An agent platform that has to prove every claim it makes before a human signs off.

FastAPIPydantic v2SQLAlchemy 2PostgreSQL 16
EDIConnect mapping interface
Generative AIData Engineering

EDIConnect

Hybrid rule-based and LLM-assisted mapping for B2B EDI documents.

PythonLLaMA 3.1Sentence embeddingsFAISS
RAG and KnowledgeGenerative AI

Netflix Q&A Assistant

A RAG chatbot that knows when to look something up exactly instead of searching for it.

LangChainChromaSentence Transformers (all-MiniLM-L6-v2)Groq
NLPGenerative AI

English to Urdu Translation

A Transformer trained from scratch on a 24,525-pair parallel corpus.

PyTorchnn.TransformerSentencePieceHugging Face tokenizers
Generative AIComputer Vision

Video Understanding Pipeline

Speech, frames, and a language model, run in parallel to summarise a video.

Whisper smallViT-GPT2 captioningPhi-3 mini 128kHugging Face Transformers
Confusion matrix of the fine-tuned ViT on CIFAR-10
Computer VisionMachine Learning

CNN vs Vision Transformer

What transfer learning buys on CIFAR-10: 69% from scratch, 97% fine-tuned.

PyTorchtorchvisiontimmViT-Tiny
The Mona Lisa, one of the three test images used for segmentation Live demo
Computer VisionMachine Learning

Lazy Snapping Segmentation

Paint a few strokes and let colour clusters separate foreground from background.

NumPyscikit-learn KMeansPillowTypeScript (demo)
Face KNN demo: a test face beside its seven nearest training faces and their distances Live demo
Computer VisionMachine Learning

Face Recognition with KNN

A from-scratch nearest-neighbour classifier on 32 x 32 faces, compared with SVM and Naive Bayes.

NumPyscikit-learnPCATypeScript (demo)
Perceptron decision boundary over two classes Live demo
Machine Learning

Rosenblatt's Perceptron

The original learning rule, implemented from scratch and animated.

NumPyMatplotlibTypeScript (demo)
Convolution versus correlation with an asymmetric kernel Live demo
Computer Vision

Convolution from Scratch

The operation behind CNNs, written by hand and compared with SciPy.

NumPyOpenCVSciPyTypeScript (demo)
Cryptarithmetic demo: column-wise search assigning digits to SEND + MORE = MONEY Live demo
Machine Learning

Cryptarithmetic Search

SEND + MORE = MONEY, solved by search, with the frontier on screen.

PythonA*DFSBFS
Bakery lock simulator: three threads, their choosing flags and ticket numbers Live demo
Systems

Bakery Lock Simulator

Lamport's mutual exclusion algorithm, one memory operation at a time.

C++pthreadsstd::atomicTypeScript (demo)
Bridge simulator: vehicles queued at a single-lane bridge with condition-variable waiters Live demo
Systems

Single-Lane Bridge

Cars, buses, a mutex, and two condition variables.

C++pthreadsCondition variablesTypeScript (demo)
Bitcoin forecasting demo: daily closing price with a chronological train and test split Live demo
Machine Learning

Bitcoin Price Forecasting

Regression baselines, stationarity tests, and walk-forward evaluation on daily BTC prices.

pandasscikit-learnstatsmodelsTypeScript (demo)
Urdu segmentation demo: an Urdu passage split into sentences with boundary triggers Live demo
NLP

Urdu Sentence Segmentation

Rule-based sentence boundaries for Urdu, where punctuation alone is not enough.

PythonregexurduhackSentencePiece
Detected tomatoes and potatoes with bounding boxes
Computer Vision

Grocery Detection and Billing

Open-vocabulary detection turns a photo of produce into a bill.

YOLO-World LRoboflow inferencesupervisionOpenCV
Training and validation curves for the CIFAR-10 CNN
Computer VisionMachine Learning

CNN for CIFAR-10

A compact Keras CNN, augmentation, and a learning-rate ablation.

TensorFlowKerasscikit-learn
Machine LearningFull-Stack

S&P 500 Forecasting App

Eight forecasting models behind one Streamlit interface.

StreamlitstatsmodelsProphetpmdarima
Data Engineering

Electronica Data Warehouse

A HYBRIDJOIN stream-relation join loading a star schema, then OLAP in SQL.

JavaMavenMySQLSQL
NLPData Engineering

Common Voice Urdu EDA

What is actually inside an Urdu speech dataset.

pandaslibrosapydubseaborn
Machine LearningData Engineering

Audio Similarity with LSH

MFCC fingerprints and MinHash banding to find similar tracks.

librosaNumPypandasMinHash LSH
Data Engineering

Instagram Engagement KPIs

Normalised engagement metrics and median baselines for fairer content comparison.

pandasNumPyTableau

Coursework and earlier projects

Smaller assignments and experiments, kept for completeness. Each links to its repository.

Everything on GitHub

Also on the resume

Projects whose code is not public. Described from the resume only.