DashR is an AI powered virtual assistant that acts a true assistant in both, physical and virtual sense.
What it does
From making you a custom beverage to helping you find the best travel deals, DashR can do it all.
-
DashARM beverage maker:
DashARM is a module which can make beverages automatically by using voice commands to control a robot arm. Our use case shows how we can make a cup of coffee while we are on our way, so a warm beverage will be ready when we arrive. All you have to do to place your order is tell Dasha what you want.
Hey Dasha, can you make me a cup of mocha?
This can be used to make other beverages and potentially also food as well.
- DashaCAM The DashCAM module allows us to view who is at the door remotely, and also send them a message if so requested. We use SMS to send an image back of the person at the door, and this uses Avaya as well as Google Cloud Platform for interaction. Not at home to receive a package? Ask Dasha to pass on instructions to the delivery man. Combining this with our IoT sensors, this subsystem can be used to make the check-in process for access-controlled buildings completely contact fee and seamless.
DashaFly
This is a travel assistant which allows you to lookup flights and allows you to get information about cheapest flights and also the covid alerts and pandemic warnings; and also a list of interesting things to do at the destination. The external functions are essentially wrappers we built for Google Places and Amadeus to accomplish this.
Watch the Demo hereDashaHVAC allows you to get information about the temperature, humidity, noise levels, light levels, and other information collected by IoT sensors, as well as control these settings using the voice assistant.
human: it's too cold in here.
dasha assistant: i've increased the temperature from 24.1 to 26
How we built it
hardware:
- servo motors
- potentiometers
- wires
- sensors
- dht11
- BME280
- 18650 batteries
- Arduino controller boards
voice assistant:
- dasha.ai (dsl)
- dsl
- nodejs
- python
- ngrok
- mongodb
- google cloud storage
Challenges we ran into
- training coherent phrases within such a short span of time
- character limit on #sayText (we actually had to shorten DashaFly's personalized itinerary list due to this)
- Argument error while training intents (thanks to the mentors for helping us figure this out) **Intents shouldn't have spaces.
Accomplishments that we're proud of
- multiple working subsystems
- working MVP
What we learned
- dasha scripting language
- language models can be complex to train
What's next for DashR
- make DashR more robust by training it on a more complex model.
Top comments (0)