Manish Devatha

Developed a question-answering system using LLM models that is trained on custom data. Using LangChain, chained together prompt templates, LLMs, agents, and new data stored as vector stores.

Analyzed historical website traffic data, including daily, weekly, and monthly patterns, to forecast future traffic using Facebook Prophet and ARIMA. This can help businesses plan server capacity, optimize marketing campaigns, and identify periods of high user engagement.

Train Deep neural network models (convolution neural networks) using Python Libraries (Keras a high-level neural networks API running on Theano) and perform image classification which resulted in automating the existing process.

Performed text mining on an unstructured dataset to identify job positions, locations & employers of the recently hired graduate students using Python & Tableau.

• Assessing a major electric company's IT and analyzing 65,000 Lotus Notes based applications
• Simplifying the IT landscape to establish a centralized governance mechanism using advanced data analytics.
• Rationalizing and Identifying modern target solutions to re-platform
• Assisting client in its technological migration from its current landscape to a futuristic target platform.
• Coordinating with multiple stakeholders across all demographics.

• Analyzed consumer insight and buying lifecycle using mobile analytics
• Performed adhoc investigative analysis of mobile activity eg cart abandonment rate, exit rates
• Created reports using funnel and analyzed the choke points for abandonment rate
• Analyzed user trends ad behaviour on the mobile app and aligned the ROI of mobile marketing investments.
• Measured all the important mobile metrics across mobile websites, mobile app platforms that contributed towards the purchase… Show more

• Analyzed consumer insight and buying lifecycle using mobile analytics
• Performed adhoc investigative analysis of mobile activity eg cart abandonment rate, exit rates
• Created reports using funnel and analyzed the choke points for abandonment rate
• Analyzed user trends ad behaviour on the mobile app and aligned the ROI of mobile marketing investments.
• Measured all the important mobile metrics across mobile websites, mobile app platforms that contributed towards the purchase lifecycle
• Redesigning UI & UX in par with customer needs and business objectives.
• Worked with global and local research teams to leverage research efforts Show less

This tricorder was designed and prototyped to fit on any standard Insulin pen and catered to various problems faced by diabetic patients such as reminding user to carry their insulin pen with them by geofencing, monitoring and prevention of multiple insulin dosage intakes, etc… using Bluetooth technology

This was designed to automate the medicine intake process, specially catering amnesia patients where they have to take multiple medications and tend to forget, this solution also provided the physician to monitor and change the dosages on the go.

Designing of an automatic accident system using two accelerometers and a gyroscope, and sending the location of the crash to the victim's relatives as well as the nearest ambulance service. This can be implemented both on four-wheelers as well as two-wheelers. The GPS location of the crash is used by an algorithm running on a web server to find out the nearest emergency services.

A self-designed remote and car prototype. The transmission is based on Manchester encoding to improve reliability of communication.

- Worked with heavily imbalanced dataset to detect if the credit card transaction is fraudulent using dimensionality reduction and outlier detection techniques.
- Used PCA and isolation forest algorithms from Scikit-learn and Pandas for data cleansing.

Created a music recommendation system that suggests songs or playlists to users based on their listening history, preferred genres, and similar user profiles using PyTorch.

Identifying if tweet was from a bot and distinguishing negative marketing. Used Natural Language Processing (NLP), Twitter APIs to gather data, and configured scalable Big Data system using Apache Spark on AWS EC2.