Nitin Parab

• Engineered and operated mission-critical database platforms supporting 99.99% uptime SLAs across large-scale enterprise environments, implementing high availability, disaster recovery, backup, replication, and operational resilience strategies
• Architected and delivered an award-winning 11-skill agentic AI system built with Claude Opus and Claude Code, integrating five operational data sources to automate Pod Migration and Decommission workflows. Delivered a unified status dashboard, real-time progress tracking, and NLP-based issue detection with resolution recommendations, reducing handoff report generation time from 15 minutes to 30 seconds per pod. Team received Salesforce Hackathon 2026 "Most Heroic" Award.
• Led APAC Pod Migration and Decommission programs, consolidating large-scale database environments from legacy data centers to AWS, migrating over 5 PB of data and delivering multi-million-dollar infrastructure savings.
• Designed and managed real-time database replication and data synchronization solutions while leading large-scale migration and modernization initiatives across on-premises and cloud environments.
• Performed database performance tuning, capacity planning, troubleshooting, and workload optimization for large-scale OLTP systems, while developing Python and SQL-based automation solutions to eliminate manual operational processes.
• Analyzed operational and infrastructure datasets to identify capacity trends, optimize resource utilization, and support strategic migration, decommissioning, and platform planning decisions at Salesforce scale.
• Served as a trusted technical advisor to engineering leadership, driving root cause analysis, platform reliability improvements, database architecture reviews, operational best practices, and migration strategy across distributed systems.

• Managed enterprise database environments supporting mission-critical engineering and business applications, delivering high availability, disaster recovery, backup and recovery, database migration, performance optimization, and operational reliability initiatives.

• Provided Oracle database administration and production support for Keppel FELS, including troubleshooting, patching, performance analysis, backup/recovery and infrastructure changes.

• Supported mission-critical database platforms for UBS Singapore's core banking and financial services applications, delivering high availability, performance optimization, database upgrades, migrations, troubleshooting, and root cause analysis within a highly regulated banking environment.

• Administered mission-critical banking database environments, providing 24×7 operational support, high availability and disaster recovery solutions, and troubleshooting complex production issues under strict SLA requirements.

• Administered large-scale database platforms supporting telecom billing, OSS/BSS, and customer management systems, delivering performance optimization, production support, and successful application releases in high-volume transactional environments.

• Managed enterprise database environments for banking, insurance, and retail clients, delivering high availability, disaster recovery, migration, backup/recovery, and performance tuning initiatives.

Key Areas: Machine Learning, Deep Learning, LLM Applications, RAG, AI Governance & Ethics, Agentic AI

Research Projects:

Old Bailey Court Records Analysis (1900–1913) | Python, Scikit-learn, XGBoost, NLP, SHAP, Causal Inference

• Built an end-to-end NLP and machine learning pipeline on 8,800+ legal case records to predict judicial outcomes from unstructured court transcripts.
• Engineered 57+ text, linguistic, sentiment, readability, and topic-based features and developed predictive models using Logistic Regression, Random Forest, SVM, and XGBoost.
• Implemented explainable AI (SHAP), fairness analysis, and causal inference techniques to evaluate factors influencing verdict and sentencing decisions.

Utility Fraud Detection | Python, Scikit-learn, XGBoost, Feature Engineering

• Developed an ensemble fraud-detection solution for electricity and gas consumption data, achieving AUC 0.792 and 65% fraud recall through advanced feature engineering and model optimization.
• Evaluated multiple supervised learning algorithms and optimized precision-recall trade-offs for highly imbalanced fraud datasets.