3. PathwayAI: A Role‑based, Real‑time AI Workspace for Admissions

" Where stories find their voice. "

Pathway AI — Collaborative AI Workspace for Admissions

Elevator pitch

Who it’s for

• Students who can’t afford admissions consulting but still want expert‑level structure and feedback
• Anyone unsure which majors/programs/schools match their interests, budget, and constraints
• Mentors/counselors/family who want to support students asynchronously in one shared place
• Teams that want real‑time presence and AI support with zero technical setup

Access and affordability (mission)

• Agency‑grade support without the agency price: AI copilots compress expert guidance into an accessible workflow
• Free, cloud‑based onboarding: invite supporters instantly; no local setup or paywalls to start
• Discovery‑first: tools help you figure out your fit (major/program/school) before you start writing
• Aid‑aware suggestions: surface fee waivers, scholarships, test‑optional policies, and value metrics early

Discovery and matching (solve “What should I study? Where should I apply?”)

• Fit Finder: interest + constraint intake (interests, budget, location, outcomes) -> suggested majors/programs/schools with rationale
• Source‑aware research: Perplexity (sonar) + curated sources expose current admission requirements, deadlines, and program strengths
• Aid & affordability filters: tuition bands, scholarship signals, in‑state/out‑of‑state differences, cost‑of‑living hints
• Portfolio alignment: map your activities/essays to program expectations so effort goes where it matters

Feature highlights

• Collaboration and presence
  • Multi‑role spaces (student, mentor, counselor, family) with clear permissions
  • Google Docs–style comments: reply, resolve, mentions; consistent glass‑UI cards
  • Designed for live cursors, conflict resolution, and offline sync
  • Invite collaborators via centered frosted‑glass modal with emoji role types
• AI writing and research
  • Gemini‑powered drafting/refinement with clean markdown output
  • CompactAI suggestions: categories, severity filters, tooltips, one‑click accept/reject, before/after previews
  • Refine panels that never auto‑replace—review and accept with confidence
  • School Agent using Perplexity (sonar) for up‑to‑date requirements and quick citations
• Editor experience
  • CoreEditor architecture: stable instance via refs; zero re‑renders during typing
  • CSS‑only mode switching for normal/focus/comment modes to protect cursor and selection
  • Auto‑save every 10 minutes with “last saved” time
  • Comment actions appear only in comment mode to reduce clutter and misclicks
• Information architecture and UI
  • Three‑column layout: requirements | writing assistant | agent/co‑pilot
  • “Agent Mode” expands the writing area while analysis runs in the right panel
  • Frosted glass design system, consistent spacing, aurora gradient chips where appropriate
  • Searchable, preloaded dropdowns (school filters); floating refine button; centered modals
• Reliability, auth, and platform
• Supabase with unified server‑side getUser() authentication (SSR)
• Cloud‑only storage and endpoints—no local DB needed for collaborators
• 60s timeouts and automatic retries for network‑sensitive tasks
• Unified error handling for predictable experiences under load

Technical barriers addressed

• Low‑friction collaboration: share a link, start working—no installs or accounts for mentors to “just help”
• Stable editing at scale: ref‑driven editor + CSS‑only mode switches prevent flicker and focus loss
• Auth consistency: SSR getUser() and unified error handling reduce intermittent failures
• Network resilience: generous timeouts and retry logic for scraping and AI APIs
• Design consistency: reusable frosted‑glass components and a strict grid/spacing system

Architecture and stack (brief)

• Frontend: Next.js (App Router), Tailwind, frosted glass design system
• Auth/data: Supabase with @supabase/ssr and unified getUser() flow
• AI: Google Gemini via @google/genai (ai.models.generateContent({ model, contents }))
• Research: Perplexity (sonar) for dynamic, source‑aware requirement checks
• Editor: Imperative CoreEditor (refs), event‑based comment markers, TipTap Marks (planned)
• Performance: Stable editor instance, autosave queues, no re‑renders during typing

My role and skills (what I bring)

• End‑to‑end product leadership
  • Translated admissions pain points into a collaboration‑first workflow with discovery before drafting
  • Drove roadmap, UX research, information architecture, and measurable success criteria
• Real‑time systems and collaboration
  • Presence‑ready architecture (live cursors, conflict resolution, offline sync)
  • Decoupled editor core from UI for future CRDT/OT without rewrites
• AI/ML integration at product quality
  • Robust Gemini integration with markdown‑clean responses, retries, and fallbacks
  • Perplexity sonar pipelines for source‑aware research and requirement extraction
• Frontend engineering at scale
  • Ref‑driven, zero‑re‑render typing loop; CSS‑only mode switches; custom events to isolate heavy logic
  • CompactAI suggestion UX for high signal and minimal flow interruption
• Platform and reliability
  • SSR auth patterns with Supabase getUser(); secure server‑side AI initialization
  • 60s timeouts, automatic retries, unified error handling; cloud‑only endpoints for easy collaboration
• Design systems and DX
  • Multi‑layer frosted‑glass component library; strict spacing and motion guidelines
  • Accessibility‑minded patterns; rapid iteration velocity with strong component reuse
• Leadership and velocity (turning the dial up)
• I operate across product, engineering, and design with uncommon range—comfortable in systems design reviews, UX critiques, or deep debugging
• I move fast without breaking quality: set standards, mentor contributors, and ship features that feel polished from day one

Current status and roadmap

• Implemented
  • Stable editor core; AI drafting/refining; autosave with last‑saved time
  • Searchable school filters; frosted glass UI components; SSR auth; longer timeouts and retries
• In progress
  • Full multi‑role presence (live cursors, selections), conflict resolution, offline sync
  • Advanced comment system (TipTap Marks, React Portal) decoupled from the editor
  • Richer CompactAI suggestion cards, category/severity filtering, batched accepts
  • Expanded School Agent capabilities and affordability‑aware Fit Finder

Call to action