Introducing Holmes

Holmes is our new receptionist and a face recognition AI system with chatbot interface. It learns to recognize people who come and go at our office, and with a little help from humans it is able to connect an new face to a name. When Holmes meets a visitor, it asks if it is OK to remember them and also asks if anyone from Emblica personnel knows who the visitor is. Next time, depending on your consent, it will remember you and greet you with a personal message. Or it might have no idea who you are, except a strange feeling that this might be someone who doesn't want to be remembered... (if you're clever, you noticed a paradox here).

How Holmes works?

Holmes tries to find any visible faces in front of it and if it sees them it sends the photo to the API to be inspected more closely.
Frames are preprocessed by cropping only facial area, that way API needs to analyze only preselected content which reduces bandwidth and latency. Holmes sends photos by using Websockets so overhead per frame will be low.

Brains of Holmes first converts the picture to numerical signature of your facial features using deep neural network. That signature is compared to other signatures Holmes knows using the same library Spotify uses for their song recommendation engine called Annoy. If match is found Holmes API takes everything it knows about you and recommends some actions based on time of the day, weather, your schedule, actions and even your facial expression, or if you haven't given consent for that it leaves you alone.

Slack chatbot

When detecting new people, Holmes is able to ask around Emblica office, if anyone knows this person. This is achieved with a Slack integration and a chatbot. In the future, we will also build more intelligent chatbots that will become parts of Holmes to augment it's capabilities. It can ask you relevant questions and you can tell it to remember certain things about you, like to show the correct public transportation timetables when he sees you leaving the office.

Contextual information personalised based on biometric signature is the future of infoscreens and dashboards.

Numerous offices have set up screens showing bus & tram schedules, lunch options or some company metrics. However, are these really equally relevant to everyone? We imagine Holmes running several dashboards and cameras around our office and offering personally relevant information based on the viewer's personal preferences. Holmes could show you lunch options from your favorite restaurants, public transportation schedules that actually get you home - and remind of you upcoming meetings.

Would Holmes or something like it increase customer interactions with your business?

Irma Savolainen
Service Design

Holmes is an Emblica Labs project. These are internal projects that we do for fun, and to discover what's possible with today's technology. Commercial possibilities or practical use cases are not requirements in Emblica Labs projects (although they can be welcome side-effects). The projects are chosen entirely based on the motivation of our employees. As the main selection criteria we use the questions: "Is this something we want to spend time on and can we learn something new from this?". Sometimes these projects end up in production, sometimes we open source them and sometimes they are left as prototypes hidden somewhere in our basement. Either way, Emblica Labs gives our employees the possibility to train themselves and discover new ways of working.

Emblica Oy
Kaisaniemenkatu 1, 00100
Helsinki, Finland


Teemu Heikkilä
+358 (0) 400 963 509 

© 2018 Emblica