DeepConv AI

The numbers are staggering. Today, nearly 1 billion people live with a mental disorder and in low-income countries, more than 75% of people with the disorder do not receive treatment. Every year, close to 3 million people die due to substance abuse. Every 40 seconds, a person dies by suicide. About 50% of mental health disorders start by the age of 14. - Ref - www.blogs.worldbank.org

The DeepConv AI is an API that can be integrated into any platform of choice, ranging from smartphones to robots. It uses a trainable billion parameter AI model that can work in multilingual. It is both empathetic as well as adaptive. The model can recommend your day, comfort in your tougher times, and much more.

If adopted globally it will impact billions of lives.

The numbers are staggering. Today, nearly 1 billion people live with a mental disorder and in low-income countries, more than 75% of people with the disorder do not receive treatment. Every year, close to 3 million people die due to substance abuse. Every 40 seconds, a person dies by suicide. About 50% of mental health disorders start by the age of 14. - Ref - www.blogs.worldbank.org

• 970 million people worldwide have a mental health or substance abuse disorder. 
• Anxiety is the most common mental illness in the world, affecting 284 million people. 
• Globally, mental illness affects more females (11.9%) than males (9.3%). 
• Major depression, anxiety, alcohol use disorders, schizophrenia, bipolar disorder, and dysthymia (persistent mild depression) were identified as leading causes of disability in the U.S. (U.S. Burden of Disease Collaborators, 2013)
• The mortality rate of those with mental disorders is significantly higher than the general population, with a median life expectancy loss of 10.1 years. (JAMA Psychiatry, 2015)
• It is estimated mental disorders are attributable to 14.3% of deaths worldwide, or approximately 8 million deaths each year. (JAMA Psychiatry, 2015)

Ref: https://www.singlecare.com/

The solution is a RESTful API for a pre-trained model that is trained on a large corpus of data that exhibits the blended skill, including the empathetic dataset, Wikipedia dataset, and conversational AI dataset. Uses transformers and seq2seq architecture to train the models. The model can be fine-tuned according to the user conversations. It is multilingual covering 300+ languages. The API is integrated with edge single board computers like the Nvidia Jetson series, Smartphones(apps), and computers(desktop applications). The model was tested with AI pioneers from the open-source community to check bias and racial/ethical problems.
Already deployed in the community with thousands of peers reviewing the model and fine-tuning it on a daily basis. The community is cross-checking the model efficiency/accuracy and optimizing the API for real-time needs.

Privacy: The conversation is an end to end encrypted and each conversation is stored locally or on the edge to secure privacy.

We are focusing on everyone that can connect to the internet, including any race/gender/ethnicity that wants to live a happy life. The solution will help to organize their life and comfort them in their tougher times. Since it covers 300+ languages it is helping a lot of communities right now.
Most of them are confused in their lives, don't know how to live a meaningful life. A mentor that understands and guides them is what they need most. Some people lost their parents, some siblings, and some close friends. They all need a hug from a godfather or godmother. This solution will watch them in every way and without compromising their privacy

The solution is currently implemented in various mobile applications, desktop applications, and web applications to interact with the user. Currently, almost one thousand people are peering the AI model to improve its performance. It is silently saving people's lives by caring for them. 
Output
1. Give a day to day recommendation basis of their previous interactions
2. Comfort and suggest solutions in mental health problems, very common issues like exam stress to losing a closed one.

We are currently using the Facebook AI Research community to test our solution(ParlAI), and currently, at most 1000 people are testing the solution. I am also an Udemy AI teacher so I just deployed it in my AI community and we are collaborated with various local android/ desktop/ web application communities to deploy our solution.

Our solution is straight to the point. No additional steps needed. Just fine tune the model and deploy according to the need.

Currently we are serving 100+ people. In one year the number will reach to 5000+ and in five years 100,000+ .

The response from customer in the long run

40+ people, including student interns and experienced professionals

We are working in Single Board Computers(SBCs) like Nvidia Jetson Nano, Raspberry Pi for almost two years and have Udemy AI teachers working in the core team. Profound knowledge in mobile-based optimization of AI models like converting TensorFlow model to TFLite, TF model to Open Vino Intermediate Representation, and the IR model blob model for DepthAI modules, and extensive research in large language models like GPT, BERT etc.

We believe in constant support and caring , we at deepflow are leveraging that policy to build a healthy community. 

Our solution is based on AI models and we are working hard to solve the problem in a holystic way. If we get the funding we will leverage it to make a better server and also fund to build a better AI model as time goes.