SDE 2 at Certa | Building GenAI Solutions with Python

Rahul Sunil • Peerlist

Grapnel, a software solution that has been developed in the Hac’KP, that is proving to match any other world-class product in the category when it was tested, can handle various requirements of the law and order organizations while working in the realm of the dark web.
Grapnel will equip police on advanced dark web search, intelligent monitoring of the dark web and dissecting it for illegal transactions and criminal activities being undertaken using the dark web. When a keyword like drug, weapon, child sex abusive material (CSAM), banking, terrorism, leaked database, etc., is entered, the Grapnel crawls in the dark web to retrieve websites having the keyword hosted in it and list them.
Images and videos having a relation with the keyword will also be listed. Also, there are options for advanced search in which users can search using a date range like in conventional search engines like Google. “The search using a keyword with the help of Grapnel is yielding a wealth of information that is otherwise unavailable in the normal web. This will help police in cracking down on illegal activities undertaken through the dark web,

Grapnel: Dark Web Monitor

• Collaborated with Microsoft, Field Data Technologies and the University of California for developing the solution.
• Developed a non-invasive method for wildlife monitoring and surveying of Endangered Species using UGV with advanced cameras and minimal motion.
• Developed Python Libraries for XBee Interfacing and Rover Controls.

Unmanned Vehicle for Wildlife Survey

BankEasy is a project that aims to cater towards contactless bank ATM/CDM transactions - a necessity that arose in the COVID pandemic, due to the extreme risk of contracting the virus via exposure.

• It involves a Flutter-based mobile application that conveniently incorporates the NFC technology in mobiles and modern credit/debit cards, alongside secure QR scanning (for payment sessions) paired up with two-factor authentication via a custom-built hardware key we call the BEKey. This, wired with Tangle networking, eliminates the use of a Credit/Debit card and provides a hassle-free, secure, contact-less banking experience to withdraw or deposit money.
• The idea conceptualization was shortlisted among the Top 15 finalists ideas of YouthTech Competition by StartAD and Abu Dhabi Department of Education and Knowledge and the prototype also won the first prize in the 'HackX' Hackathon hosted by NITC.

BankEasy

An accident is an unpredictable and unintentional event. Prompt assistance to people involved in vehicle accidents could make a significant difference in the consequences of such accidents and has a high possibility of saving one's life. 

This system aims at providing early detection of accidents by detecting the turbulence on the vehicle and communicating the information immediately to the emergency responses on time to provide quick assistance for the injured person.

We presented our Solution in Springer's 3rd International Conference on Computer Networks, Big Data and IoT ICCBI 2020 and got it published as a conference paper.

Accident Alert System

smarTangle is a Tangle network-based smart end-to-end Saas Food Warehouse Optimization system solution, with a primary focus on raw food materials. The project covers the management of all the stakeholders in a food chain right from the farmers with the produce, to the Warehouse, the subsidiary stores and finally the customers under one digital decentralised ledger umbrella.

• It houses the tracking of the movement, quality and shelf life of the produce while also forecasting the regional requirement of raw materials based on daily demands and other seasonality factors in the subsequent stores for a period of up to 10 weeks using the FbProphet Time Series forecasting model.
• This brings about a food currency system that keeps a footprint of all food transactions from when its produced to when it's finally in the hands of a customer. Hence, ensuring transparency in the system for both the warehouse and the customer - immensely reducing the panic buying and food wastage factors, something that is currently not in motion for raw food materials, unlike cold storage.

This project helped me in developing a complete SaaS solution which consisted of developing a responsive Django-based Web Application for the Warehouse End and a Flutter-based mobile application for the customer end. A SQL based database was used for Web End while a NoSQL based database was used for the Mobile End. The Time-Series forecasting algorithm incorporates improved efforts from an ARIMA model to make use of the FBProphet based on the dataset. The transactions are recorded in a public Tangle (COMNET) network and is hosted in the GCP compute Engine.

smarTangle

A Django based Website for booking halls in our college. The features in the web app include:

• Created an app that allows users to login via google sign in and book halls either on the basis of event dates or on the basis of the hall required for their event
• They can see and manage their existing bookings and also cancel them.
• The superusers verify the bookings made by the users through email or the app itself.
• Available halls including all the details (capacity, availability time, location, etc. ) can also be viewed.
• Admin can add and remove users and halls.
• Build using Django, HTML, Bootstrap, Google API for authentication and authorization (using Auth 2.0)

ZaRoc - Hall Booking Site

Using Python module Scrapy, the data from the URL entered is crawled through based on the given keywords and flags those which supports offensive crimes (cyberbullying, pornography, child and women abuse etc). 

This is done by saving the sentences, which contain certain keywords, in a CSV file and is then updated using sentimental analysis by checking whether the lines are offensive or not.  The updated CSV file, containing the URL, offensive sentences, and frequency of keyword, can be manually verified by the admin. 

A portal providing provision for the government is created, to view the reported sites and also for the public to report sites. These websites are then observed and necessary action is taken accordingly.

DefenceCrawler

Our idea is to bring up a user friendly, secure app for storage of different medical records, prescriptions and other relevant reports of a patient where he/she can upload them to their account in the mobile application. This will help them get easy access to their reports when required. The prescriptions will help users to get their required medicines and also can help doctors to know about the medicines taken earlier and their medical history. 
The app includes additional features - daily progress of steps and distance covered, calories burned, medication reminders  etc. 


BitMedi

Environmental conditions such as extreme rainfall, earthquakes, landslides and floods often cause natural disasters which lead to tremendous loss of life and property, causing great disruption in people’s lives and the economy. In 2018, across the world there were 315 natural disaster events recorded with 11,804 deaths, over 68 millions of people were affected, and 131.7 billion dollars in economic losses.

After a disaster strikes, governments and private organizations engage in reconstruction efforts of infrastructure, such as roads, bridges, power and railway lines, houses etc. This is a very labor and capital-intensive task. Moreover, doing this at a disaster site, with its multiple associated risks and challenging terrain, poses an additional challenge.
Motivated by this scenario, in order to help the needy victims in the affected areas, this edition of e-Yantra Robotics Competition (eYRC 2019-20) presents the theme “Construct-O-Bot”.

In this theme, the arena is an abstraction of a disaster site where the robot picks the construction material and traverses paths in order to deposit it at the site to be reconstructed. In order to maneuver over these paths, the Construct-O-Bot has to use intelligent line-following and path-planning algorithms to reach safely and quickly using shortest paths.
After reaching the site, the Construct-O-Bot carefully has to place the material at the required positions which may include placing the material at different heights from the ground. It has to deposit all required material at multiple construction sites, navigating through various terrains.

ConstructOBot - Eyantra

Using an AI-powered model embedded in a drone which can detect the infections or symptoms in plants at an early stage using Tensorflow. It could thus prevent the outbreak of the disease and reduce the intensity of crop failure to a great extent. Thereby saving a large area of crops by eradicating the infected part.