Shreya Maity

Client A - Mirai Science Co. Ltd.

• Improved pest prediction accuracy by 15% using XGBoost regression and time-series models on weather and pest datasets.

• Built scalable forecasting models for multiple stations, achieving R² > 0.85.

• Automated ETL pipelines using Python, reducing preprocessing time by 30%.

• Developed spatial risk heatmaps using Matplotlib, Seaborn, and Cartopy.

• Integrated YOLOv4-based pest detection with prediction pipelines.

Client B: Avia Corporation

• Developed an AI-powered Chrome extension using JavaScript and REST APIs, reducing recruiter screening time by 25%.

• Built a Job Matching System using TF-IDF, BERT embeddings, and vector similarity search.

• Developed the backend using Python, Supabase, and Streamlit, achieving over 80% matching accuracy.

Client C - Softturns Technologies Inc.

• Built an automated lead generation pipeline using Apify to extract business data from Google Maps.

• Automated workflows using n8n, integrating a Telegram bot and Google Sheets.

• Developed social media automation using the Meta Graph API for Facebook and Instagram posting.

• Reduced manual effort and improved scalability through end-to-end automation.

• Developed a house price prediction model using machine learning.

• Built an employee attrition prediction system using XGBoost.

• Created an NLP-based sentiment analysis system.

• Working on an object detection, i.e., image annotation using YOLOv4, and trained YOLOv4 tiny models using the Darknet framework based on CNN.
• Fine-tuned model parameters by making changes in epochs, and optimized training pipelines for improved accuracy and performance.
• Convert the YoloV4 Tiny model into TensorFlow Lite so it can easily be deployed with the help of Android.
• Conducted rigorous testing and evaluation of models for real-world deployment.
• Implemented text extraction from images using optical character recognition (OCR) techniques.