Implemented Retrieval-Augmented Generation (RAG) pipelines using LangChain, OpenAI APIs, and vector databases (Pinecone, FAISS) to improve enterprise query accuracy by 35%. Built agentic AI workflows with autonomous agents for retrieval, summarization, and decision support, reducing manual research time by 50% and boosting operational agility. Engineered scalable ETL/ELT pipelines with PySpark, Apache Airflow, and AWS Glue, automating ingestion and transformation of 5TB+ data daily across distributed sources. Operationalized ML models on AWS SageMaker with Docker, Jenkins, and GitHub Actions, reducing deployment cycles by 40% and ensuring reproducible pipelines. Designed LLMOps pipelines with MLflow, Prometheus, and Grafana for fine-tuning, experiment tracking, and drift monitoring, cutting model degradation by 25%. Applied prompt engineering and embedding optimization to improve retrieval quality, achieving 90% relevance in enterprise assistant queries. Created Tableau dashboards and

Machine Learning Engineer – Real-Time AI Systems at Vinjmauri Lab (Brain-Machine Interfaces), UMBC

January 1, 2024 - July 1, 2024

Architected a CNN-Transformer model fusing real-time EEG and facial data to achieve 97% classification accuracy with sub-10ms inference time, enabling real-time neurotech applications. Developed EmoFormer, an optimized vision transformer that reduced inference costs by 30% through pruning while setting new SOTA benchmarks on FER2013 (+8%) and AffectNet (+5%). Delivered an end-to-end real-time neuro-inference pipeline in Python, integrating ML models into production for a commercial partner and achieving 81% accuracy under a strict 200ms SLA, meeting mission-critical latency requirements.

Data Scientist at Infosys

June 1, 2020 - July 1, 2023

Built and standardized data processing workflows using Python, SQL, and R, reducing data preparation time by 35% and accelerating client project delivery. Developed and deployed AI/ML models across deep learning, NLP, and computer vision using TensorFlow, PyTorch, and Scikit-learn, improving churn prediction by 18% and demand forecasting by 22%. Designed high-performance data pipelines using Apache Spark and Hadoop, enabling real-time processing of 1M+ records per day for enterprise clients. Created and optimized hyperparameter tuning frameworks with Optuna, reducing training and experimentation time by 20%. Integrated cloud-native AI solutions using Azure Data Factory, AWS EMR, and Snowflake, ensuring scalability and reliability across hybrid infrastructures. Established MLOps frameworks with Kubernetes and MLflow to automate deployment, retraining, and governance—reducing downtime and manual intervention by 40%. Mentored junior data analysts on model validation, CI/CD practices, an