Shivam Gulati

Architected a production agentic real estate search platform with multi-agent orchestration in LangGraph/LangChain, improving multilingual search (voice-to-search) outcomes and increasing search session depth by ~30% while reducing resolution time by ~40%. Built forecasting and valuation models that reduced price estimate error by ~30% and streamlined multimodal content operations, cutting product

Designed and deployed end-to-end AI/ML systems on GCP (Vertex AI, BigQuery, Vertex AI Feature Store) for a leading international bank, with lifecycle monitoring including drift detection, SHAP-based auditability, and automated alerting for production degradation. Led architecture discovery with banking product teams, translating business requirements into scalable ML system designs documented via

Developed and productionized an LLM-based classification pipeline that clustered 80K+ job requisitions into ~50 categories, reducing manual review and saving significant annual FTE hours using Azure ML/SageMaker with MLflow tracking. Built fairness-audited NLP demographic inference for ~500K applicants with >90% balanced accuracy and demographic parity validation, plus bias detection/fairness audi

Built regression-based failure prediction and Cox Proportional Hazards RUL models across 13 power plants in Azure ML Studio, enabling risk-ranked maintenance scheduling. Reduced unplanned downtime by ~20% through improved maintenance decisions.

Enhanced data quality for 5M+ e-commerce listings using NLP, text analytics, and ML pipelines, reducing user-reported catalog bugs by ~5% (250K+ listings remediated).

Completed a Bachelor of Technology (B.Tech.) in Computer Science at ITM University, Gurugram in 2014.