Kavyasri Jadala

Dubai, UAE

• Designed and deployed a demand-forecasting platform across 4,000+ SKUs and 33 stores for a large Philippine retail conglomerate, using global and hierarchical LightGBM/XGBoost models.
• Served SKU-store forecasts with confidence and model metadata through FastAPI; automated ingestion, retraining, validation and promotion with Airflow, MLflow and Docker.
• Built an AI application-development platform turning requirements, decks, screenshots and Figma designs into deployable full-stack apps via schema-first domain modelling, multi-agent code generation, automated validation and human-in-the-loop governance.
• Built geospatial network planning combining forecasts, trade areas, mobility data, H3 indexing and PostGIS to rank expansion sites and quantify cannibalisation.

New Jersey, USA

• Architected an enterprise knowledge graph integrating 20+ commercial systems on AWS Neptune with RDF, OWL, SHACL and governed semantic models.
• Built a Graph-RAG assistant translating business questions into governed, citation-supported graph queries, replacing ~80% of recurring ad-hoc data requests.
• Engineered deterministic and probabilistic entity resolution across CRM, offline engagement and web analytics with confidence scoring and survivorship.
• Built segmentation and customer-lifetime-value models contributing to a 15% improvement in sales conversion; built an in-house voice-analytics pipeline projected to save ~$600K a year.

West Lafayette, IN, USA

• Built a capacitated vehicle-routing model with time windows for a 300-vehicle J.B. Hunt fleet, cutting peak-period misrouted shipments 20% (~$300K a year).
• Built shelf-intelligence vision with YOLOv8 and Faster R-CNN for planogram compliance at Meijer.
• Developed Transformer and wav2vec2 audio-text alignment for under-represented languages (~90% accuracy); presented at INFORMS Analytics 2024.

Dubai, UAE

• Reduced daily delivery-demand forecasting MAPE from 25% to 18% across e-commerce, grocery and quick-commerce.
• Owned demand, capacity and cost forecasting for a $40M+ annual logistics budget across the UAE, Saudi Arabia and Egypt.
• Embedded model scores into operational APIs and workflows; designed A/B tests for routing and delivery features; led a team of 21.

Gurgaon, India

• Built inventory-optimisation and demand-forecasting models for 800+ products, reducing days inventory outstanding from ~300 to 120 (~$250K a year).

Gujarat, India

• Integrated sensor and maintenance data into predictive-maintenance capability that raised plant uptime by 20%.