Point Cloud Segmentation for Bathymetric Lidar- Worked on bathymetric lidar point cloud segmentation and achieved 95%+ accuracy for various locations, wrote various optimisations to reduce prediction time by 67% and memory footprint by 73% for training.- Developed an active learning system to continually improve the model for new data. Continiously engaged with client to improve the system according to their needs.Implementation of Digital twins for PV plantso Collaborated with other team members to come up with an algorithm for nominal data extractionwhich confirmed that having a digital twin can reduce power wastage by around 1-3%.o Extracted nominal data, suggested and implemented algorithms for data cleaning. Used plots toreason my algorithm and improved based on the team feedback.o Implemented VAEs and other Feed Forward networks to clean the data and do drift detection.Fault prediction for PV plantso Worked on an extremely hard problem with label imbalance of around 300:1.o Implemented generic functions for meaningful plots and training models.

Crescer AI builds robust, scalable, and cost-effective AI for renewable energy and operations applications. I identify strong opportunities for partnerships with players in the solar energy space to advance their O&M and analytics capabilities using AI.

Took classes of Machine Learning for juniors and introduced them to the topic.Maintained the AI wing repository for OpenCode to improve Open Souce culture in college

- Selected as XROS fellow for GmetriXR.- Used openai APIs to interact with GPT-3.5 model to automatically generate metaverse using the information provided by client. Wrote various scraping scripts to get additional information from URIs.- Worked with FastAPI and SQLAlchemy to expose endpoints to get the necessary information in the required format to generate the Metaverse. Explored prompt engineering to get better output from openai.

- Majorly worked with data team to fine tune various open source models and use them for DPG projects.- Worked on some interesting projects like recommendation system for fertilizers using metarank, GandhiJi voice clone using eleven labs, chat bot for kerala farmers using audio from krishi support phone calls.

● Open source contributor for Mango Market and contributing to solcpp, a c++ sdk for solana. The sdkallows C++ users to connect with solana blockchain and use web3.js functionality in C++.● Implemented websocket using beast library in C++. Used promises and futures to maintain concurrencybetween threads. Allows users to connect to solana websocket and call a callback whenever data change.● Ported all the websocket functions like onAccountChange, onLogs and many more from web3.js to solcpp

Managed events online for freshers like Consilio.Helped managing webinar for Machine Learning.

Created an end to end machine learning platform to train and deploy model.Project focuses on enabling users with less knowledge of Machine Learning to quickly build and deploy a ML model.Publicly available government dataset is used to train the models.Implemented building and deploying ML models automatically.

Learnt about open source contribution and development.Worked with my pod member to develop a Weather app.

Developed a full stack website for a college based startup. Website dealt with managing and providing coolers to students at a low price.Website contained an admin side to manage orders and a client side to place the orders.