Meeting Minutes (Meeting with TA)
Date: 2026-01-28
Time: 09:40 - 10:25
Location/Platform: In-person(KHRA L1-350)
Facilitator: Aditya
Attendance
| Name | Present (✅/❌) | Notes |
|---|---|---|
| Aditya | ✅ | |
| Pragya | ✅ | |
| Prajin | ✅ | |
| Smriti | ❌ | |
| Mustafa | ✅ | |
| Ashwin | ✅ | |
| Salaar | ❌ |
Discussion & Agenda
User Stories
- Need 40+ US
- breakdown US further(too broad User Stories)
Wireframes
- Wireframe should cover a US, and label each wireframe with its associated US
Tech Stack
- use FAST API instead of django
- django is overkill, consider using django rest framework
- fast api also has an admin panel
AI, Form Filling, Transcription
- need a lot of audio data to test llm models, but NDA issue
- emergent behavior, small models dont have that, need to finetune
Other
- aws has serverless lambda functions(service spins up only when needed to save cost and resources)
- UML:
- clear up lines in UML
- use right angles lines
- put label closer to entity
- Sequence diagram:
- gone too wide, ahs all entities from all interactions
- documenting main interactions
- has too much broader info
- else, its good
Decisions Made/Next Steps
- client questions:
- gpt-4o mini, ask about privacy
- Use Azure speech to text instead of openai whisper
- calculate cost, and ask client is he is okay for paying that
- openai api should be good(they dont store any voice/texts)
- to optimize system, we need audio data, forms; ask about NDA
- tech stack:
- switch to FAST API
- look into serverless(lambda functions)
References/Attachments(optional)
- [Link to Documents]