• Spearheaded the development of a STEM career recommendation system using LLM, integrating Direct Preference Optimization (DPO), resulting in 20% improvement over accuracy on highly personalized and accurate career suggestions.• Implemented Retrieval-Augmented Generation (RAG) pipeline and deployed the model on AWS SageMaker, ensuring scalable, robust performance in handling extensive user interactions and data processing.• Integrated LangChain as a framework for ReAct agents, employing Tree of Thought reasoning and enabling multi-turn dialog interactions to optimize coordination between the recommendation engine and external data sources, enhancing the system's responsiveness and efficiency.

• Developed data-driven and physics-informed neural network (PINN) methodologies to identifying center manifold for nonlinear dynamical systems, significantly reducing model characterization & computation costs by 30%.

NASA AEROFUSION-MLUQ: Mars Lander Aerodynamic Data Fusion using Machine Learning with Uncertainty Quantification• Developed Mars Lander Aerodynamics Multifidelity Multiphysics Simulation on Ansys Fluent. Created geometry in python and mesh in ICEM CFD by automation script. Decreased workflow time by 80% earlier used to take a day now in 10 min.• Implemented Machine Learning to predict trajectory of Mars lander, a project sponsored by NASA Langley Research Center.• Ran Airfoil WMLES simulations & created custom Python scripts to generate publication-quality plots for diverse quantities.• Expertly orchestrated High-Performance Computing (HPC) job submissions, leveraging SLURM scripting to streamline simulation processes. Demonstrated proficiency in Linux environments, enabling efficient management and optimization of computational resources.

• Pioneered a machine learning-based Agricultural Yield Prediction System, integrating diverse datasets (soil quality, weather, crop types) to enhance crop prediction accuracy by 30% and irrigation accuracy by 25% using advanced neural networks and LSTM models.• Led the integration of MLOps practices for seamless deployment of yield prediction models on AWS, automating training, validation, and CI/CD pipelines. Conducted A/B testing on deployment strategies to ensure robust, scalable performance in production environments.• Identified key agricultural segments using KMeans clustering & formulated optimized resource allocation strategies, increasing crop yield by 15%.• Predicted potential yield drop for each segment & enhanced crop stability by 20% by analyzing & addressing issues in sensor & environmental data.

• Developed an AI-driven diagnostic tool, integrating Neural Network with SPO2-based thermal imaging, achieving 95% accuracy in COVID-19 detection, aiding in early diagnosis during the pandemic.• Orchestrated creation of innovative healthcare platform with bio-inspired machine learning model for multi-disease prediction, significantly improving patient-doctor interaction and healthcare outcomes by leveraging key features extracted from semi-structured medical data.

• Built and optimized data pipelines, automating ETL processes for data cleaning and feature engineering, while enhancing SQL queries in Apache Spark to decrease latency by 60%, resulting in a 20% increase in data processing efficiency and a 40% improvement in query performance for over 250GB of customer data.• Automated data collection and reporting processes, reducing manual data entry by 40%, and developed real-time Tableau dashboards, enhancing decision-making efficiency by 30% and supporting customer retention strategies.• Boosted operational and ML model development efficiency by 25% through automated exploratory data analysis, contributing to more effective customer churn reduction and retention initiatives.

• Reduced materials cost by $2,000 per device by analyzing thermal performance data and optimizing cooling systems for electronic enclosures.• Decreased project delivery timelines by 15% by integrating simulation data into cooling strategies for custom power supplies, improving workflows.• Enhanced product reliability by 20% in extreme conditions through rigorous data analysis and integration of environmental test standards into thedesign verification process