I build AI pipelines, web backends, data tools, and automation — and deploy them on AWS and Azure.

I built a conversational AI assistant directly into my portfolio site. Recruiters and visitors can ask anything about my projects, skills, and background — and get grounded, cited answers in real time, powered by RAG (Retrieval-Augmented Generation) and Llama 3.3 70B via Groq.

I built a convolutional neural network that classifies food images into 101 categories using two-stage transfer learning with MobileNetV2. Upload any food photo and get instant top-3 predictions with confidence scores — live on Hugging Face Spaces.

Stacked LSTM for stock trend pattern detection, served via Amazon SageMaker Real-Time Endpoint. Flask frontend on Azure App Service — multi-cloud MLOps. Pattern detection only, not financial advice.

A full ML pipeline benchmarking TF-IDF + Logistic Regression, a Keras LSTM, and Claude Haiku via Amazon Bedrock on 50,000 IMDb reviews. The winning model is deployed live as a serverless REST API on AWS Lambda with a Flask frontend.

1970s Boston house price predictor. Linear regression on 13 features with log-transformed target (test r²=0.74). Live slider UI on Azure; full analysis in rendered notebook.

Unsupervised segmentation of 7,043 Telco customers into five behavioural archetypes using K-Means and Hierarchical Clustering. Input a customer profile to receive a segment label, a plain-English placement explanation, and a dynamically generated retention strategy.

Churn prediction with plain-language risk explanations a retention team can act on. Logistic Regression wins (ROC-AUC 0.8362), benchmarked against Amazon Bedrock Claude Haiku 4.5 zero-shot — tied at 70%. Flask API on Azure.

SMS spam detection comparing three approaches — Naive Bayes, SVM, and Amazon Bedrock (Claude Haiku zero-shot) — on the same test set. Every prediction comes with a plain-English explanation of the words that drove it. Class imbalance addressed explicitly. Deployed on Azure App Service.

Stack Overflow lost 97.7% of its post volume since 2016 — and its momentum now anti-correlates with hiring demand. This project proves the signal is broken, then builds a four-source replacement index to answer which languages to hire for.

Full-stack Flask portfolio with a custom admin dashboard, PostgreSQL, a REST API, an AI assistant powered by RAG, Cloudflare R2 media storage, and CI/CD deployment to Railway via GitHub Actions.

Monitors a stock's daily price change and fires WhatsApp alerts with the top 3 news headlines via Twilio whenever the move exceeds a configurable threshold.

Log workouts to Google Sheets using plain English. A local LLM parses your session, calories are calculated via the MET formula, and each exercise is written as a timestamped row — no manual entry.

Draw a digit on the canvas — the CNN predicts it in real time. Trained on 60,000 MNIST images. Confidence score and probability bar chart across all 10 digits. Live on Azure App Service.

PDF document summarizer using Amazon Bedrock (Claude Haiku) with production LLMOps patterns: versioned YAML prompt templates, automated ROUGE evaluation, Bedrock Guardrails for PII redaction, and MLflow tracking of every inference call. Deployed on Azure App Service.

T5-small fine-tuned on the AESLC corpus (14k Enron email/subject pairs) to generate professional subject lines from an email body. Five suggestions returned at varying temperature levels — from beam-search conservative to top-p creative sampling.