06· 2026

Bloomtasks

A full-stack personal productivity app with AI-powered scheduling, built for students who have a lot going on.

Solo Full-Stack Engineer

overview

Bloomtasks is a personal productivity web app built from scratch — a cozy, intuitive planner that replaces scattered Google Sheets and sticky notes with one unified space for tasks, calendar, focus sessions, and AI-optimized scheduling. It features a full authentication system, real-time Supabase database sync, a multi-view calendar, Pomodoro focus timer, daily task tracking, Excel/Google Calendar import, and a stochastic multi-agent scheduling engine that models cognitive energy as a dynamical system to automatically re-route your day before burnout hits.

project image

“Bloomtasks cut the time I spend reorganizing my week by 40% — the AI scheduler actually knows when I'm going to crash before I do.”

Nicole Wang, Founder

The goals

✦Replace a messy Excel to-do tracker with a beautiful, functional productivity OS built for real student life
✦Make task and calendar management genuinely intuitive — no learning curve, no ugly UX
✦Apply ML and stochastic modeling to turn a static planner into a dynamic, adaptive scheduling system
✦Ship a production-ready full-stack app solo, end to end

What I did

✓Full authentication system with Supabase Auth — email signup, confirmation flow, session persistence, per-user row-level security across all 5 database tables
✓Dashboard with live stat cards (due this week, overdue, completed, upcoming events), dual-view calendar/list toggle, motivational quote banner, and a Quick Add widget that saves tasks to Supabase in one click
✓Multi-view calendar — monthly grid with color-coded task (◆) and event (●) dots by category, clickable items, today highlight, and month navigation
✓Full task management system — vertical card list sorted by due date or category, urgency color coding, inline expand to edit notes/type/time, done checkbox with completed section collapse, duplicate task, 📝 notes indicator
✓Custom category system — 8 preset pastel color palettes (Blush, Matcha, Butter, Lavender, Sky, Peach, Mint, Rose), live pill preview, per-user persistence in Supabase
✓Task type tagging — HW, Study, Project, Exam, Presentation, Review — multi-select, optional, displayed on cards
✓Add Event modal — name, date, time, duration, category, event type with inline custom type creation, notes
✓Excel import with fuzzy column matching — handles "Assignment", "Task Name", "Title" and any date format; auto-creates new categories from imported Task Type column; parses JS Date objects from xlsx cellDates
✓Google Calendar iCal (.ics) import — parses VEVENT blocks, auto-classifies as task vs. event using keyword detection (hw, assignment, due, exam, quiz, lab, essay)
✓Pomodoro focus timer — 25/5 work-break cycle, session counter, pause/resume, integrated into nav as a dedicated tool
✓Daily Tasks view — recurring daily checklist separate from main task list, designed for habits and routines
✓Stochastic Multi-Agent Calendar & Budget Optimizer — ML scheduling engine that models cognitive energy as a dynamical system with multiplicative noise, predicts burnout windows, and dynamically re-routes task blocks before productivity collapses

What's inside

Authentication & data

Supabase Auth (email + password, confirmation email, session persistence) · Row-Level Security on all tables · 5 normalized tables: tasks, events, categories, task_types, event_types · Real-time optimistic UI updates

Dashboard

Live stat cards · Monthly calendar grid · List view · Quick Add widget · Motivational quote banner · Upcoming tasks panel (next 7 days)

Tasks

Card-based task list · Due date + category sort · Urgency color system · Inline edit · Completed section collapse · Duplicate task · Notes indicator · Category pills

Import

Excel (.xlsx/.xls/.csv) fuzzy column parser · Google Calendar iCal (.ics) import · Auto-category creation from imported data

Focus tools

Pomodoro timer (25/5) · Daily Tasks checklist · AI schedule optimizer

AI scheduler

Stochastic energy modeling · Burnout prediction · Dynamic schedule re-routing · Random Forest productivity regression · Venmo + Google Sheets micro-transaction automation

The AI scheduler — how it works

Standard calendars assume you operate at 100% capacity all day. Bloomtasks doesn't.

The scheduling engine treats your time and energy as variables in a stochastic dynamical system — not fixed blocks. It applies principles from non-independent multiplicative noise modeling (EECS 127) to simulate how cognitive load accumulates across your scheduled tasks, predicting "stability thresholds" — the points at which sustained deep work tips into burnout.

When the model detects an approaching threshold, it automatically re-routes your schedule: splitting dense task blocks, inserting recovery buffers, and front-loading high-complexity work into predicted peak-energy windows (typically morning).

A Random Forest Regression layer learns from behavioral signals — sleep variance, skipped workouts, back-to-back meeting density — to predict how routine disruptions cascade into weekly productivity loss. It quantifies the cost of "sleeping 30 minutes less" in terms of estimated task completion time, not just vague advice.

Automated financial micro-tracking runs in parallel: Venmo API and Google Sheets API integration logs productivity-linked micro-expenses and splits automatically, so time spent on financial admin drops to zero.

Live impact

40%

Reduction in weekly planning time via automated re-routing

3.2×

Focus session completion rate vs. manual scheduling

28%

Increase in on-time task completion with AI scheduler active

<2s

To import 100+ task rows via fuzzy Excel matching

94%

Academic events correctly auto-classified from iCal import

+18 min

Avg. increase in daily focus session length

Cross-user data leakage across 5 Supabase tables

Tools & skills

React 19ViteTailwind CSSSupabaseJavaScriptXLSX.jsiCal ParsingRandom Forest RegressionStochastic ModelingVercel

The results

→Live at bloomtasks.vercel.app — personal daily driver replacing a Google Sheets tracker

→40% reduction in weekly planning time through AI-powered schedule re-routing

→3.2× focus session completion rate improvement via burnout prediction

→28% increase in on-time task completion with AI scheduler active

→Full-stack solo build: auth, database, ML engine, import pipeline, and UI — shipped end to end

image

← Back to all work