I build mobile apps and tools for AI evaluation.

I’m a first-year student at the University of Toronto, intending Mathematics + Computer Science.

Recently, I built the Apple Watch client for Providence at Hack the North. I’m also exploring how to test what agents actually do when tool calls fail or evidence is incomplete.

yusef.syed@mail.utoronto.ca

Selected work

An evaluation harness that compares an agent’s proposed actions with the effects its tools actually produced, including partial failures and missing evidence.

Deterministic checks, resumable runs, and explicit unknown outcomes.

Python · Inspect · SQLite · Docker

Personal experiment infrastructure

Providence

Hack the North 2026

A voice-controlled Mac assistant built with my team. I built the native Apple Watch client: hold-to-talk capture, request routing, and spoken-result playback.

Request IDs keep approvals and results attached to the right interaction.

Swift · SwiftUI · watchOS

My contribution: Apple Watch client

An annotation-only garment-completion study, with calibration and missing-context tests. Methods and limits are documented.

Agent Proof

Developer tools

A TypeScript CLI for running reviewer-selected checks and saving redacted evidence. Runs from source; not a sandbox.

ShiftProof

Agent workflows

A local scheduling prototype with independent constraint checks and coordinator approval. Uses synthetic data.

A synthetic WebMCP inbox where an agent prepares a plan and the owner decides. No messaging backend or send capability.

Research methods, results, and limitations

Product demos

Three short walkthroughs. Each includes notes on what is released, illustrative, or built with sample data.

Aesthetics AI

Released on the App Store

An iOS fitness product that turns physique and meal photos into structured reports, training plans, and progress tools.

About this walkthrough

Product walkthrough for the released iOS app. The report, plan, meal, and progress examples are illustrative and labeled in the video.

React Native · Expo · Supabase · Structured LLM API · RevenueCat

  1. Video input and a structured report
  2. Training plan and editable meal estimates
  3. Progress, coach dashboard, and App Store handoff

CallReclaim

Sample-data prototype

A missed-call recovery prototype with consent-aware follow-up logic, lead capture, and an owner dashboard; this walkthrough uses sample data.

About this walkthrough

This walkthrough uses sample data. Real-client messaging remains gated by sender, carrier, routing, and acceptance-test checks. The case study explains the implementation.

Next.js · Supabase · Twilio Voice/SMS · Structured LLM API

  1. Missed-call event and approved text-back
  2. Lead qualification and conversation state
  3. Owner dashboard and follow-up action

Tiraz

Released on the App Store

A released iOS wardrobe product for outfit feedback, digital closet organization, and saved looks.

About this walkthrough

The app is publicly available on the App Store. This video is an illustrative concept visualization, not a live screen recording.

React Native · Expo · Supabase · Edge Functions

  1. Outfit-feedback concept
  2. Digital closet and saved looks
  3. Subscription and account-flow design

Open source

CPU/CUDA incomplete-gamma derivatives, with author-run local validation. This contribution has not been merged.

Validation notes
Full contribution record

Contact

I’m interested in supervised research during the school year and software or research engineering internships for Summer 2027.

yusef.syed@mail.utoronto.ca

Available full-time May–August 2027. U.S.–Canadian dual citizen; open to relocation and remote work.