Lyntra · AI Product · Design from 0 to 1
Lyntra is an AI-powered schoolwork planner that helps students break down complex assignments into actionable plans to help them meet their deadlines.
As Design Lead, I drove the product from 0 to 1, defining the AI planning workflow, core interactions, and feature priorities through launch. I then translated analytics and student feedback into the V2 design's product structure.
Final Experience
Plan → Review AI steps → Edit → Schedule
To make AI-generated work easier to understand and trust, I introduced a dedicated Plan section that separates assignments needing review from those already scheduled. Each plan is broken into editable steps, helping students understand the workload, adjust the schedule, and stay in control before adding it to their calendar.
Ask Lyntra → Choose or add an assignment → Confirm details → Generate plan
Rather than asking students to navigate a broad set of AI tools, Ask Lyntra guides them through a specific academic goal. I shaped the conversation logic with engineering, connecting it to Canvas coursework and adding a confirmation step so students can correct the AI’s understanding before it generates a plan.
Task details → Start focus mode → Complete step → Finish the task
I connected task details directly to focus mode, turning a larger assignment into a clear sequence of manageable steps. Progress, pause, resume, and completion states help students focus on the current action while staying aware of how much work remains.
The Challenge
Newly generated plans appeared alongside daily tasks, making it difficult for students to tell what was new, what had synced, and what still required review.
The chatbot offered a broad set of tools and templates, but provided limited guidance for turning a specific assignment into a structured, reliable plan.
Main tasks, subtasks, study blocks, and focus steps used overlapping structures, obscuring how an assignment connected to the next actionable step.
Design Decisions
In V1, Home carried new tasks, daily priorities, weekly planning, and progress in one place. For V2, I separated these responsibilities across Today, Plan, Calendar, and Ask Lyntra. This gave AI generated work a predictable place to appear and kept each surface focused on one primary user question.
The V1 chatbot offered a flexible set of AI tools, but students had to decide how to structure their requests and provide the right context. For V2, I organized the conversation around specific academic goals, connected it to synced coursework, and added a confirmation step before plan generation. This made the interaction more guided while preserving students’ control over what the AI understood.
V1 used a flexible task and subtask model that reflected a general project management system. In practice, main tasks, subtasks, study blocks, and focus steps created overlapping levels. For V2, I simplified the model around how students understand academic work: one assignment contains a finite sequence of steps, with one step surfaced as the current action and connected directly to Focus Mode.
Reflection
A useful AI response is not the end of the experience. Lyntra taught me to consider what users need to understand, decide, and do after the AI generates a plan.
Automation only feels helpful when users know what happened and can correct it. Clear system states, confirmation, and editable steps became essential parts of building trust.
In a startup, speed matters, but making a product public is not the end of the design process. Shipping a focused V1 helped us learn from real behavior quickly, while continuous iteration turned those insights into a more connected V2.
Streamlined practicum matching, reducing admin work by 94% and task time by 78%.
Reframed the site's information architecture, reducing task time by 30% and driving 10+ inquiries.
Owned a real-time truck performance dashboard, improving user satisfaction by 25%.