Projects

Product experiments, applied AI systems, and tools I built across the last few years.

deepwell app icon

deepwell

Live on the App Store2026

A calm home for deep work. I built deepwell around a hope I genuinely hold: that long, uninterrupted stretches of attention are worth protecting, and worth keeping. You do not start a timer, you begin a block of one to four hours, and when it ends you commit it with a note, the way a programmer commits code. Those blocks quietly grow into months of green squares, streaks, and a monthly letter written back to you about how you worked.

  • One stats engine on every platform. It is written in Rust, compiled to WebAssembly on the web, and mirrored natively on iOS, so every streak, record, and heatmap square matches to the minute.
  • Interruptions are treated as a fact of life. The block pauses with you, asks gently what pulled you away, and gives your full focus time back when you return.
  • A monthly letter grounded in your real notes and rhythms, plus Siri, the Action Button, Live Activities, widgets, and Apple Health integration.
  • One account everywhere. Blocks sync through Postgres with row-level security, and anything committed offline merges cleanly when you reconnect.
deepwell app screenshot 1
deepwell app screenshot 2
Tiny Rafeeq (Caregiver Platform - Playdate Connecting) cover

Tiny Rafeeq (Caregiver Platform - Playdate Connecting)

2025

A caregiver-facing app that helps children connect through playdates while tracking early childhood milestones.

  • - Designed a matching system that pairs families by location, child age, and temperament fit for seamless playdate planning
  • - Integrated weekly wrap-ups so parents can monitor growth and receive gentle prompts
  • - Accessible User Interface, feeling approachable to caregivers with diverse needs, minimalistic
Open Knowledge Graph
Wavelength (Spotify Mood App) cover

Wavelength (Spotify Mood App)

2025

A music app, using the Spotify API, that connects people through shared listening habits and generates mood-based playlists from their top tracks.

  • - Built a playlist generator that adapts to your mood using your most-played songs.
  • - Designed a matching feature that helps users discover others with similar music tastes
  • - Awarded Honourable Mention at the MADDATA 24-Hour Hackathon, recognised for its unique social purpose.
Open Knowledge Graph
Digital Twin for Government Official cover

Digital Twin for Government Official

2025

Built a personalized AI clone to help a government official streamline tasks and engage more effectively with stakeholders.

  • - Designed a digital twin trained on the official's past articles, interviews, and conversations to capture his voice and expertise.
  • - Conducted in-depth interviews and iterative feedback loops to align the system with his working style.
  • - Delivered a Delphi+Python backend, now actively used by the official as part of his workflow.
Open Knowledge Graph

University Admissions Advisor (Rule-based Neuro-Symbolic Tool)

2025

A guidance platform that blends human expertise and symbolic reasoning with an LLM to recommend university programs worldwide

  • - Combined admissions data and knowledge vaults with neuro-symbolic AI to suggest programs aligned with a student's profile
  • - Matched users not only on admission likelihood but also on personal interests, and global match
  • - Built as a transparent decision applet, showing students why certain programs are and aren't a good match
Open Knowledge Graph
Project Pinguinus cover

Project Pinguinus

2022–2024

A digital screening tool that helps caregivers identify potential developmental disabilities through simple questionnaires.

  • - Created a questionnaire-based app that flags developmental concerns and suggests clear next steps
  • - Designed with a caregiver-first approach, making medical guidance more accessible outside clinical settings
  • - Clean accessibility-first UI as well as multi-language support, that offered structured insights advised by professionals in the field
Open Knowledge Graph

F1 Predictive Racing Telemetry Dashboard

In Progress

A real-time analytics platform backed by Formula 1 telemetry, designed to predict race outcomes and provide insights into driver performance.

  • - Training machine learning models on an open racing dataset to forecast lap times, pit stop strategies and tire performance.
  • - Focusing on predictive analytics, turning race data into insights that mirror professional F1 race engineering tools.
  • - Integrating a what if simulator, allowing users to adjust parameters like tire choice or fuel load and see projected race outcomes
Open Knowledge Graph

The Doodler

A simple whiteboard to sketch and save quick doodles.

Wall of Doodles

Saved doodles stay in this browser only.