Project : AI BenchProject overview: NLP platform for developing various Generative AI plugins and training LLM models.Team size : 5 members• Implemented a LLM model which understands audit and product manual documents ,and brings out page wise search resultswhich helps business stakeholders to take decisions on a timely basis which resulted in reducing documentation time by 50%.• Integrated Flow wise, lang flow into the platform which helped in creating 12+ LLM apps for different audit teams.• Developed Vectorization technique using Chroma for different use cases in the platform, increased accuracy by 30%.• Led the team to track and optimize the AWS Sagemaker resources using GPUtil, which reduced 40% of cost on resources.

Project:HR Absenteeism : (Predicting unplanned absence in car manufacturing plants to aid in manpower to forecast number ofemployees needed to operate on a work day - Greenplum, ML, Snowflake [6 team members])• Implemented statistical time series models like SARIMAX, SPLINE, HIGH to predict the unplanned absence happening in eachassembly team in five car manufacturing plants. The ML models predict across teams, shifts and overall plant wise.• Developed data pipelines for model retraining and predictions were stored in data lake. Thus resulting in automating theprocess and reduced manual process by 70%.• Weekly predictions are executed and models get evaluated with metrics such as MAE, MSE on team wise and plant wise.• Continuous development on model performance like longitudinal analysis, rolling average error rate, feature engineeringwhich resulted in increasing model accuracy from 82% to 94%.• Implemented Offline MLops architecture along with the data engineering team wherein pipelines run once a week.

Smart Electricity consumption and forecasting in Textile Industry using IoT sensors which collect power data and with the help of ML model deployed in the cloud it predicts and forecasts the power consumption which is displayed on the mobile app.

Coach Assist : (Cricket smart wearable devices project - IoT, ML, Cloud Deployment, App Development [8 team members])• Created a Machine Learning model to forecast bat trajectory using Unity’s DOTS physics engine. Created a strategy forback-testing used for simulating cricket shots in Unity engine which improved testing efficiency by 90%.• Derived displacement algorithm by applying feature engineering on acceleration features from the data.• Predicted bat speed, wagon wheel, shot impact pitch points, shot type using the rotation and velocity features from thedata(MySQL) received from IoT devices.• Created a Machine Learning model to predict bat lift angle of cricket shots using Unity’s vector system.• Developed a Shot classifier model in Unity based on the Quaternions and Displacement values from the data.Player Assist : (Cricket smart wearable devices project - ML, IoT, Data analytics [12 team members])• Increased throughput for Rest API Integration and SQL queries by 60% on reducing the payloads.• Derived and developed Data Visualization plugin for the mobile app which resulted in an increase in usage of the app.• Derived data insights from MySQL DB for mobile apps which helped in producing Dashboards for players’ analytics.• Done Sentiment Analysis on players’ feedback survey for b2b app which helped in improving UX by 50% .• Improvised efficiency of data transfer (CSV) from IoT devices to mobile application by implementing data manipulation/cleaning which resulted in reducing 95% of noise. Used Regular Expressions to manage data anomalies.

• Implemented player movement and level design for a mobile game.• Developed a back-end system for a client on firebase and MySQLite.