As Thomas Edison said, 'If we did all the things that we are capable of, we would literally astound ourselves'. I continue to try and be as diverse in my projects to build projects I'm proud off.
Locomotive Combo Optimizer
An entire product suite that looks at the combo's owned by the organization to perform analytics and suggest replacement parts when it comes for repair. The tool also help standardize the communication between the supply-chain and shop-floor teams and inturn contributed to the $2M service program.
Reusable Chatbot Framework
The goal of the project was to build a centralized chatbot framework for the multiple implementations of chatbots across the organization. It used open source libraries and tools such as RasaNLU, SkLearn, spaCy and tensorflow. It quickly became a widely adopted tool for data-retrieval in the analytics driven organization. Also built a few framework plugins to allow services such Amazon Lex, Microsft LUIS and Google Dialogflow to be available via the framework.
Analyise, Trace and Monitor Application (ATMA)
A click-stream analytics tool with a library to help calculate the ROI of any application. This was done by using the data generated by the library relating to customer segmentation, product adoption and effective product management. This application helped track the process flow expected to be followed v/s actually followed while calculating the benefits in $ of how valuable one product was compared to the other.
CI/CD Upgrade - Containers
Proposed a change in CI/CD pipeline within the organization after evaluating the impact and value. Recommended a change from a simple AWS EC2 architecture, to a container (Docker) based architecture with an orchestration system (Kubernetes). The proposed solution was expect to bring a ~$200k instant benefits. Also ideated and performed proof of concepts on using serverless architectures for our web applications.
Trelleborg
Manufacturing companies often have the issue when a part gets discontinued or when it's out of stock. Often there are alternate parts available that can perfrom the exact same purpose. Since this information is dynamic and constantly changing it requires re-training technitions on regular intervals. As a solution, formulated and developed the 'Alternate Part Suggestion' tool that used tensorflow to train on the existing parts and their dimensions to recommend alternate parts from the available intventory that they can use.
In my opinion, one of the largest challenges in IoT is the constant need for human intervention to be able to make some decisions. For example - if we had to automatically turn on our microwave to heat food, the microwave would not know at what temperature to do so. A human would need to intervene, search online and provide that as an input. Can the microwave not directly talk to the internet? This project tried to challenge that phenomenon by using web syntactics and symentics to allow devices to talk to themselves with the correct information and thereby require as minimal human intervention as possible.
Open-Peer Review System in SciBase
The first of it’s kind Open Peer Review System (OPRS) of SciBase was an open and transparent framework that allowed the research commuity to engage in positive scientific disucssion/critique about exisitng research work in the scientific domain. Users could host their work independently or choose to discuss exisiting content fetching from Internet Archive’s non-profit library.
Step-wise navigation for the blind
Different online map services are very useful for people who can be fully aware of their surroundings but not so much for people who are blind - especially because of the uncertainity in the GPS location. This system used a unique mathematical triangulating algorithm that learned the exact step-wise map by learning the walking pattern of the blind person along with the relevant google map information. This was done by placing a few sensors on the stick which helped them walk and would vibrate in the direction they needed to turn.
Real-time health analysis of a plant
An IoT solutions using Arduino microcontroller to suggest farmers the health of their crops and suggested recommendations to obtain a successful yeild. It took into account the weather variations, soil fertilify, light available, air quality, etc.
