Sushant Shambharkar

Architected and deployed multi-agent wealth management platform for autonomous customer servicing, portfolio intelligence, and investment advisory.

• Built planner-executor-reviewer agents with persistent memory, real-time market data retrieval, and ML-driven risk profiling
• Autonomously resolved customer queries, generated portfolio rebalancing strategies, and recommended personalized investment actions
• Improved engagement by 35% and increased portfolio diversification
• Reduced advisor response latency by 60%

Upgraded real-time ad delivery systems using edge ML models and autonomous optimization pipelines.

• Improved CTR by 30% through dynamic ad inventory optimization
• Increased viewer retention by 20% via low-latency personalization
• Built autonomous optimization pipelines on PyTorch, Kubernetes, Snowflake, and Delta Lake

Architected NLP2SQL platform using fine-tuned LLMs to automate dashboard query generation.

• Improved dashboard creation efficiency by 200%
• Reduced dashboard load times by 3x through automated aggregation framework
• Led design of automated aggregation framework that optimized query execution

Designed self-healing experimentation infrastructure and scalable user analytics pipelines within Google Cloud.

• Embedded ML-based flag conflict detection and rollback prediction in Mendel experimentation tool
• Reduced code-induced failures by 25% and minimized hotfix cycles
• Built scalable user analytics pipelines and automated A/B test orchestration
• Achieved 7% boost in long-term retention and 10% in short-term engagement via behavioral clustering

Architected real-time NLP platform for extracting behavioral signals from client communications.

• Built real-time NLP platform using Apache Flink and transformer models
• Created multi-task ranking models for urgency, sentiment, and response prioritization
• Deployed retrainable APIs with automated evaluation pipelines for continuous model optimization
• Reduced response latency and improved service consistency

Built ML-driven workforce optimization systems for route planning and task duration prediction.

• Improved field productivity and reduced operational downtime through ML-driven optimization
• Led migration of training and inference pipelines from SAS to PySpark
• Delivered 10x reduction in processing latency while eliminating licensing costs