Ajinkya Pol

• Object Detection & Analytics for Agricultural Yield Optimization and Operational Efficiency

Led the development of an object detection solution using UAV imagery (RGB, NIR, Red band) to detect mature and immature pineapples across 26,000+ hectares.
• Implemented EfficientDet-based detection with Reinforcement Learning for bounding box refinement, improving localization accuracy in complex field environments.
• Translated detection results into actionable insights for labor allocation, yield forecasting, and supply chain optimization, reducing manual effort and improving operational efficiency..

• LangGraph Agent – Conversational AI for Agricultural Data Insights

Developed an intelligent agent enabling natural language interaction with multispectral UAV data for crop health, temporal analysis, and yield forecasting.
• Structured agricultural data into knowledge graphs using LangGraph to support explainable, context-aware insights.
• Extended the Pineapple Object Detection solution by providing an intuitive interface for agronomists and operations teams to query field analysis, enhancing decision-making and operational efficiency.

• Warranty Clustering & Suspect Claim Analysis for Manufacturing Operations

Developed advanced data-driven solutions to cluster warranty claims and identify suspect claims in manufacturing, improving quality control and reducing financial risk.

• Applied clustering and statistical feature engineering to group similar claims and identify anomalies.

• Built models to flag high-risk claims, improving investigation efficiency and reducing warranty costs.

• Enabled data-driven insights for quality control and operational decision-making.

• Harvesting Intelligence: Optimizing Tea Quality and Yield with Reinforcement Learning

Developed a precision harvesting system using reinforcement learning to optimize harvesting periods based on weather and other factors, increasing tea yield by 35% and quality by 20%.

• Achieved 30% reduction in operational costs through adaptive, cost-efficient harvesting strategies.

• Ensured robust performance under varying weather conditions, improving sustainability and consistency.

• Advanced Traceability in Agriculture: A Hyperspectral ML Approach

Implemented hyperspectral data analysis techniques to achieve granular and precise product traceability throughout tea manufacturing process.

• Developed and deployed algorithms for the accurate quantification of disease percentage, coarse percentage and moisture content enabling rapid intervention and reducing the risk of contaminated or compromised products.

• Algorithms use: PCA, SVM, XGBoost.

• Predictive Modeling for Disease Risk in Tea Estate Sectors

Demonstrated a reduction in disease management costs through targeted interventions. By focusing resources on high-risk sectors, the project led to a 18% decrease in overall disease management expenses.

• Provided tea estate managers with actionable insights to optimize resource allocation. By anticipating disease probabilities, resources such as pesticides and manpower are allocated more efficiently, contributing to cost savings.

• Algorithms used: Gradient Boosting