AI Meets YOU

Apply

Applications open

AI Incubation Program

Designed to help creating

SCALABLE BUSINESS SOLUTIONS

powered by state of the art technology tightening academia-industry collaboration

0R8A1869

1/7

Team

VAZGEN
HAKOBJANIAN

Founder and CEO at Magical Labs
Partner at SmartGateVC, Investor
16 years in AI and ML
Co-founder at Teamable
Board Member of YerevaNN

ARSEN
YEGHIZARYAN

Scientific Researcher
Ph.D. in Theoretical and Mathematical Physics, YerPhI
Co-organizer of Datafest Yerevan
Machine Learning Team Lead at Webb Fontaine
CEO at Portmind

Program

MAY 16 - MAY 30,

2022

Showcase/ Demo Day

DEC 23 - DEC 28

JAN 10 - MAY 15,

2022

NOV 22 - DEC 23

Applications/
Selection

Research/

18 weeks

Onboarding/

Matching

The representation of the results of the incubation program at the end of the 25 th week. Assuring to have primarily technical and scientific circles of Demo Day invitees for advanced technical discussions of the research.

Call for applications and info sessions in relevant educational institutions. Selection of 3-4 students by each supervisor depending on their area of expertise and students’ strengths in the field.

Two-week-long personalized learning plans for students. Matching the supervisors (along with their selected students) with startups/research groups/companies based on relevant areas of expertise.

AI models refinement and testing by students throughout the following 18 weeks of continuous mettings with startups/research groups/companies.

Alumni

researchers

graduated

companies

patentable technologies

letters of intent signed with AUA

SUPERVISORS

Batch 2

FormSpotter FormSpotter provides consolidated fitness resources all-in-one platform with a personalized workout routine and live feedback on form with an assessment of their form. Achievement By using image recognition from cameras, the company wanted to use pose estimation libraries to layout the body skeleton to then track the movements on the body to see how well or correct someone is working out. The research team used several libraries and approaches to achieve the goal.	Foodstruct Foodstruct is a set of nutrition tools popular in Google. It has a Diet analyzer that is also number one on Google and popular among a lot of people. Achievement The diet analyzer needed input on what the user ate during the day. The fastest way of doing this was using a Camera and some image processing to recognize the food(s). The research team developed an algorithm that can recognize food from a photo for 101 food types and accomplished it with 99.19% accuracy.	IMED APPS LLC IMED APPS helps plastic surgeons and cosmetologists to manage their media content (before-after gallery) in an easy, fast, and intelligent way. It is a SaaS-based B2B platform for Aesthetic Specialists. Achievement To reach the goal, the team needed an object detection model, which they developed with 93% accuracy. The company wanted to carry on the research after the program to improve the model.
Amaros Medical Amaros AI is a Software as a Medical Device (SaMD) that empowers eye care providers and life science companies to detect, track, and predict disease progression. Achievement Amaros AI worked on disease diagnosis and disease progression prediction tasks and wanted to focus on image enhancement using deep learning. During the research, the team could substitute math with AI, which is noticeably faster, adaptive, and has the potential to substantially improve.	3dzook 3dzook is a creative team of AI engineers and illustrators who have created the AILA - AI algorithm which learnt to draw illustrated portraits. 3dzook wanted to use AI in the illustration generation process also in real-time video conversion using GANs. Throughout the research, the team concentrated specifically on creating a new algorithm that produces more realistic/hand-drawn results while keeping the dzook style and they had noticeable success in real-time video conversion.	Rumors LLC Rumors Monitoring was founded at the beginning of 2015. The team has developed an online automated platform that aims to ease the media monitoring and analysis processes. Achievement The company wanted to use deep learning and especially neural networks for better and more accurate data extraction from web pages. The research group approached the challenge from two main angles: computer vision (CV) and natural language processing (NLP).
4U.am 4u.am is one of the fastest-growing Armenian startups. It is an online gift shop specializing in choosing the right gift for the right person. Achievement 4u.am wanted to create a recommendation system to automate the gift selection process with artificial intelligence. To achieve this goal, the company needed to collect its own data correctly, which the supervisors helped set up. They have developed four different models so that the company can test and improve.	Dowork.ai Dowork.ai proposes a data-driven, AI-powered solution that takes human language readable requirements as an input and as an output generates project work breakdown and estimation for each task. Achievement Dowork wanted to train a neural network that can generate work breakdown structure and predict estimates for each task based on project requirements.The research team conducted many experiments on the challenge. As the company lacked its own data, the model was trained on other datasets.	Expper Technologies Expper Technologies is transforming the childcare industry by building a companion robot Robin to help hospitals to provide social-emotional support to children and as a result, reduce the cost of Child-Life programs and increase reimbursement size up to 30%. Achievement Expper Technologies wanted to develop a dialog system for children-robot interaction. the research team conducted various experiments using the GPT-2 model.
RoboDel Technologies RoboDel Technologies is making an autonomous delivery robot which will use bike-lanes/sidewalks for transportation. The robot is designed for the food delivery industry and will work in both platform-to-customer (food aggregator) and restaurant-to-customer models. Achievement The research team developed a segmentation model that allowed detecting various road components from an image. The existing model had low accuracy in specific cases and need more data and AI model experimentation.	ForgeFiction ForgeFiction is a platform that allows communities of fiction fans to create original stories and fictional worlds together. Achievement The company wanted to build various models: an engine for book recommendations that would suggest similar books to users’ preferred ones; generation of book and chapter synopsis based on the chapter text; categorizing users based on writing styles and writing level.	Frismos Frismos is a mobile gaming company developing and publishing casual games. Achievement The company wanted to have a prediction on purchasers/non-purchasers to show ads more efficiently, also they wanted to do user segmentation (based on monetization). The research team used various approaches to predicting churn rates from the company’s tabular data. However, the challenges were broad and need more time to yield deployable results.

Batch 1

3dzook The startup uses an AI algorithm to create characters and artworks to increase the productivity of artists and designers in different fields. AI Incubator Achievements A semantic segmentation algorithm able to determine different parts of the face aimed at boosting their business process and mobile application.	CoinStats Top cryptocurrency portfolio tracker that provides live prices for Bitcoin, Ethereum & 5000 altcoins. AI Incubator Achievements Data scraping, work on the development of an AI algorithm to predict cryptocurrency price and identification of good/bad traders.	Velvioo The startup uses generative adversarial networks (GANs) to create virtual influencers by generating faces and adding them to real photos with the help of photoshop. A face-generation algorithm based on GAN (Generative Adversarial Network) aimed at boosting their marketing results. AI Incubator Achievements A semantic segmentation algorithm able to determine different parts of the face aimed at boosting their business process and mobile application.
EasyDMARC Software as a service to implement the most powerful and effective email authentication system, called DMARC, to block email scams and phishing. AI Incubator Achievements A model, which predicts whether a given request is a phishing attack or not with 95% accuracy. Observation of anomalies and application of time series outlier detection techniques to inform about possible phishing attacks.	RetentionForce A B2B SaaS product with personalized marketing automation and customer care solution for small and medium-sized businesses using chatbots. AI Incubator Achievements Prediction of churn using both supervised and unsupervised classification and regression techniques.	Embry Tech A technology to turn any shoes into biometric data tracking and wellness monitoring device. The only hardware+software solution that allows 100% consistency with weight tracking, along with other physical activity tracking. AI Incubator Achievements Identification of the activities, during which weight measurements are the most or least precise by saving time and effort for future data collection. Also, a model, which explains considerable variance in weight.
Expper Technologies A B2B robotics company that creates robots for corporate businesses. Robin the Robot helps businesses optimize expenses without the involvement of human resources, ensuring 24/7 customer service. AI Incubator Achievements An emotion-generation algorithm and a vision on how to implement it in the future.