Headquarters SWE Internship


Nov 24 - Nov 30 : 

Designed the messaging/chat database schema

Implemented 2-way SMS using AWS SNS + Lambda and saved inbound messages to the database.

Built outbound SMS delivery logging using CloudWatch Logs Subscription Filters + Lambda → sent results to our server.

Designed the frontend chat UI (iMessage style) and implemented message list/detail/compose pages.

Implemented routing logic for patient SMS replies depending on available support team members.

Added real-time chat using SSE and completed end-to-end chat flow.


Nov 17 - Nov 23:

Implemented ghost preview for rehab templates and refactored related APIs.

Rebuilt the z-index system with Matan to unify multiple conflicting layers.

Added small UX improvements (email/phone input on patient creation, map UI for in-person appointments, timeline fixes).

Fixed multiple bugs and improved consistency across the interface.

Participated in bug-hunting and tested login + PWA features.


Nov 9 - Nov 16 : 

Added edit functionality on the timeline for appointments and evaluations (easy time adjustments).

Fixed SCAT-6 PDF auto-population and enabled sending the completed PDF via email.

Implemented saved priority/order for rehab items, preserving layout across refresh and syncing with the patient view.

Added UX improvements :

“Go back to today” badge + smooth jump animation

Patient sidebar auto-selects the first valid patient when filters apply

Tooltips showing evaluation completion time

Completed 20+ tasks on the Monkey Board and made 50+ commits.

Joined engineering meetings and appreciated Lucas’s updates and Orestis’s guidance.



Nov 3 - Nov 8 : 

  • Availability page - made users make appointment on availability page. Reused booking component.
  • Recommendation lists are shown on plan-builder. Clinicians can select recommended re
  • Timeline(Planbuilder) : Users can add rehab-items on the timeline. Timeline has zoom feature, and users can resize rehab-items and control time span. 
    • Evaluation icons : Evaluation is kind of survey that help diagnose users' status. Drafts are saved and can resume from where users stopped.
    • Appointment : users can make appointment. Appointments are also shown as icons. 
       Technical challenge : Timezone issue. Saving in db, applying org timezone, or even patient timezone. Time discrepancy between local and production environment


Oct 27 - Nov 1 : 

  • Launched and improved the new timeline: Gathered clinician feedback (Holly, Mary, Austin), replaced evaluation bars with icons, added appointment tab, hover details, and refined zoom/new-item visibility.
  • Refactored and fixed bugs: Improved API efficiency for fetching evaluations/appointments; solved complex resizing and alignment issues; completed 50+ commits and 10+ action items on Monkey Board.
  • Collaborated effectively: Productive meetings with clinicians and the engineering team; discussed API reuse, HIPAA compliance, and new productivity tools like Cursor 2.0.
  • Reflections & next steps: Learned lessons from accidental API deletion, suggested group brainstorm for AI rehab-recommendation feature, continued developing domain knowledge and communication through clinician interactions.
Oct 20 - Oct 25:  

  • Enhanced PlanBuilder UX: Added a drag-based “add-new” popover like Google Calendar with rehab item search; created appointment/evaluation chips with hover lists and smooth zoom animations; improved rehab-bar movement and snapping behavior.
  • Implemented and optimized key features: Built expandable popover modals, optimized “delete all,” “revert,” and “template” actions, and refined SQL/UX logic for replacing evaluation bars with icons.
  • Debugging and collaboration: Fixed Google Calendar webhook bug, debugged environment-specific popover issue, collaborated with Orestis via whiteboard and pair coding to resolve logic errors.
  • Growth and learning: Reflected on API and UX efficiency, explored AI agent integration in Masters project, and appreciated internship extension through May while preparing for continued innovation and teamwork.

Oct 13 - Oct 18:  

  • Focused on new PlanBuilder development: Continued major 2-week project with over 50 commits and full implementation of requirements; collaborated closely with Lucas, Orestis, and clinicians for design and usability feedback.
  • Applied feedback quickly: Added recurring icons and implemented user feedback promptly, emphasizing fast iteration and customer satisfaction as key strengths.
  • Collaboration and teamwork: Incorporated valuable feedback from Mike on timezone and optimistic updates; welcomed Matan to the team and benefited from his insights.
  • Reflections and learning: Improved communication to avoid deployment confusion, deepened understanding of React, TypeScript, and GraphQL, and identified the need to clarify requirements and ask more questions proactively.



Oct 6 - Oct 11 : 

  • Survey System Expansion:
    Added new survey questions for Headstrong, Travis One AG, and SISC in the database — a time-consuming but crucial task to improve data collection and user input flow.

  • Feature Polishing — CEUs & Plan Builder v3:
    Fixed bugs and built missing mobile pages for CEUs, and started developing Timeline (Plan Builder v3) to strengthen core scheduling and visualization features.

  • Collaborative Debugging & Knowledge Sharing:
    Joined a productive Friday debugging session where engineers reviewed each other’s code and discovered hidden issues together.

  • UX & Branding Enhancement:
    Implemented organization logo display on the booking page via S3 integration, improving both functionality and brand visibility.

  • Performance Insights & Growth Mindset:
    Investigated app performance bottlenecks (S3 latency, dev-mode overhead) and reflected on skill growth, communication, and long-term improvement as a software engineer.



Sep29 - Oct 5 : 

During the week of September 29 – October 5, 2025, work focused on stabilizing and automating the Scheduler system. The booking interface was revamped by replacing the old “dashboard” with a new BookingList page, adding filters for clinicians, event types, languages, and time zones to improve usability. A Zustand store was introduced to manage shared states such as organization ID, timezone, and scheduler state. Several bugs were resolved, including duplicate Google Calendar events, incorrect created_by values, and recurring availability issues. On the backend, automated webhook renewal was implemented using AWS EventBridge and Lambda (executed every 3 hours), along with a reminder notification system that sends SMS and email reminders 24 hours, 1 hour, and 1 minute before appointments. The participant editing flow and UI components were also refined, and all updates were successfully deployed to production.








Sep 22 - Sep 28 : 

Scheduler Building

  • Booking & Participant Pages

    • Refactored booking page to support different user types (patient, team member with/without portal, public), enabling them to add patients, caretakers, and associations across tables.

    • Added dropdown lists for patients and team members with portal permission; integrated query parameter injection into scheduler workflow (using zustand store).

    • Fixed booking page bugs, including the “go-back” button error and participant page issues.

    • Removed unused API file (book/create), unused store selectors, and extracted shared interfaces (BookingParams, Patient, Caretaker, TeamMember) into types/scheduler.ts.

  • Payments & UX

    • Implemented Apple Pay and Google Pay, updated their card order, and deployed to production.

    • Fixed UI/UX issues: carousel empty card bug, availability modal popup position, and timezone/state priority logic in scheduler welcome & booking pages.

  • Infrastructure & Notifications

    • Prevented Slack notifications from firing when running locally.

    • Implemented SQS + Lambda to queue Google Calendar webhook notifications, resolving DB connection pool issues.

    • Now, all Google Calendar changes are reflected in our DB within ~1 minute.


Webhook & Calendar Integration

  • Registered domain with Google to watch calendar changes.

  • Built flow:

    1. Google sends push notification → API receives headers.

    2. Headers stored in DB (google_calendar_channels).

    3. Events fetched & processed to update appointments.

  • Fixed major blocker: DB pool exhaustion caused by concurrent pushes. Solved by routing through SQS FIFO + Lambda consumer.

  • Deployed consume API route for Lambda → our server processing.



Sep15 - Sep 21 : 

-New pages: Built Payment, Address, and Insurance pages, focusing on UX by benchmarking other services and designing data schemas aligned with FHIR standards.
-Clinician scheduling: Polished the clinician scheduling page, incorporating 20+ pieces of feedback from boss, which significantly improved both design and practicality.
-Booking page: Collaborated with boss to improve the booking page design and resolved a time zone issue that caused incorrect date display on test devices.
-Stripe integration: Added environment variables in Amplify, enabling Stripe API to run successfully with the live key.


Seo 12(Fri) - Sep 14(Sun) : 

Features & Improvements

  • Added buffer_inputs and min_advance_minutes fields on the clinician availability page

  • Introduced a location (address) page for patients to enter details after booking, with Google API autocomplete

  • Moved the “Add” button to the bottom of the caretaker section

  • Enhanced calendar UX: improved drag functionality with precise grid alignment, fixed hover line alignment, and corrected z-index issues in the availability modal

Notifications & Calendar

  • Fixed Google Calendar invitation issues in production

  • Enabled Google Calendar email notifications for appointment invites

  • Resolved JSON parsing error causing intermittent POST failures in the process-notification API

  • Fixed timezone issue that caused Slack and SMS messages to display one day late

API

  • Added a public scheduler address API endpoint: /api/scheduler/public/address/route.ts





Sep 8(Mon) - Sep 11(Thu) : 

  • Infrastructure & Deployment

    • Integrated Amplify ENV and successfully merged/deployed to the main branch.

    • Fixed time-related errors in the deployed environment by eliminating improper Date() usage.

  • Integrations & Notifications

    • Completed Google Calendar, SMS, and Slack integration fixes by adjusting API call patterns.

    • Added AWS SNS SMS integration and stored Google event/meet links in the Appointment table.

    • Separated DB transactions from external notifications (Slack, SMS, GCal) to improve booking UX speed.

  • Database & Schema Enhancements

    • Created clinician–patient relation table when new patients are added.

    • Added buffer_minutes and min_advance_minutes columns to the schedule table.

    • Introduced a refresh token column in the organization table for GCal organizer use.

  • Frontend & User Experience

    • Improved booking pages:

      • Disabled/grayed out unavailable dates.

      • Enabled filtering slots by clinician ID and query params.

      • Combined First/Last Name fields into one input and made Name + Email mandatory.

    • Added caretaker input fixes and a confirmation page after booking.

    • Built a shared scheduling component for public, simple, and advanced booking flows.

  • Code Quality & Collaboration

    • Rebasing and syncing with main branch.

    • Addressed duplicated environment variables for Google credentials.

    • PRs submitted for fixes and new features, with ongoing feedback incorporated 



    • Sep 4(Thu) - Sep 5(Fri) : 

    Clinicians can make mark their availability (included 'recurring' function as well) 
    Worked on Patient Side Appointment Page
    Based on the availability clinician made, patients can make reservation.
    Made it one-click UI so patients with concussion can easily use it 
    Attached Google Calendar Invitation, so participants of appointment could get google calendar invitation
    ( To Do : Mobile view , Calendar Page that includes stakeholders' schedule ) 



    Sep 3(WED) - fixed db schema(reflecting convention), designed FE/BE of Clinician Side of Scheduling 
    (To Do : showing occupied time table on front end -> tomorrow )
    (To Do : finishing rest of scheduling dashboard -> tomorrow ) 

    Sep 2(TUE) - designed db schema for new feature (scheduling system) 



    Comments

    Post a Comment

    Popular posts from this blog

    2025 Spring CS686 DevOps - Final Project

     Final Project Overview -make MSA service -use EKS -Automate -deploy blue-green strategy  -deployment flow : Nightly Build -> QA -> UAT -> Prod  Steps 1. Make MSA codes  Dismantled existing code into three services - backend1, backend2, and frontend 2. By Git Action test and deploy Made GitAction Workflow('Nightly-Build') to start temporary EC2 and test(smoke-test) the code. Used dockers to run three different services and tested if it works. 3. Deploy to QA Server Using GitAction, deployed it on QA server(single EC2 server). Now we can see it at IP address of EC2 instance.  4. Make EKS Cluster and Deploy Blue Service of UAT  Make VPC and EKS Cluster.  Make Node Group. Need to use good enough instance.  Using Git Action and yaml file from Infra repo, apply configuration to running pods on EKS.  Now your microservice is deployed on EKS and you can access it through url.  5. Switch it to Green Service of UAT  Deployment of Gr...
    Read more

    Open Source Project - Open Energy Dashboard

    Image
      Open Source Project - Open Energy Dashboard Introduction  During last 4 weeks our CTI team (Hyunoh, Abas, Ezequiel, and Mon) worked on Open Energy Dashboard open-source project. Open Energy Dashboard(OED) is a project where it is used as an energy dashboard so that all the users can manage the energy usage of their worksite. ( link ) The Issue  We as a team coded a line group reading test case(c11) in JavaScript and TypeScript using Docker and Visual Studio Code, ensuring values are read correctly to reflect 14 days giving hourly point and middle readings. ( Link )   Test cases are important to enhance accuracy of a software. By writing and updating test case, we estimate we improved test coverage by 30% and enhanced data accuracy by 25% in a Dockerized environment, ensuring seamless integration and reliable energy value processing.  To resolve the issue, our team 1. looked up for development document,  2.starting Docker and activate web and database...
    Read more