Bishwadip Maitra
Software Engineer & AI for Enviornment Researcher
My Journey
Professional Experience
Working to implement Low Cost Sensor (IOT) networks across Bangladesh for PM2.5 measurement. Creating Data Pipelines and Data Science (ML DL Models) for PM2.5 modeling and data collection.
- Developed deep learning pipelines for calibrating air quality sensor data with high accuracy.
- Conducted spatial analysis to identify pollution hotspots and support decision-making.
- Translated complex ML outputs into actionable environmental insights.
- Advised on experimental design for health risk modeling using sensor + satellite data.
- Translated 30+ business requirements into sprint-ready tasks, ensuring zero missed deadlines across 3 Agile cycles.
- Improved user conversion by implementing E-NACH flows, vernacular support, and co-applicant features.
- Reduced onboarding failures by 28% through comprehensive testing and edge-case handling.
- Accelerated feature delivery by 35% via workflow automation and cross-functional coordination.
- Built a React-based admin dashboard, reducing manual data processing time by 50%.
- Increased team productivity with reusable components and scalable UI architecture.
- Boosted development velocity by 30% using Git workflows and AI-assisted coding.
- Reduced code review cycles from 3 days to 1 day through cleaner code practices.
My Work
Projects I'm proud of
- Architected a microservices-based system with REST APIs and Kafka event streaming for real-time and asynchronous communication.
- Built AI-driven coaching using GPT integration to deliver personalized workout insights and injury prevention feedback.
- Implemented Spring Cloud (API Gateway, Config Server, Eureka) for scalability, service discovery, and centralized config.
- Enabled intelligent activity tracking and context-aware recommendations based on user behavior patterns.
- Reduced API response time by 82% using Redis caching and optimized database indexing.
- Ensured data consistency with pessimistic locking, preventing double-booking in concurrent scenarios.
- Secured the platform with JWT-based authentication and robust input validation against SQL injection.
- Designed scalable REST APIs with pagination and efficient query handling for large datasets.
- Built a voice-enabled system to automatically classify and route complaints using NLP techniques.
- Improved grievance handling efficiency by mapping user inputs to relevant government departments.
- Designed for scalability and real-world governance use cases under national-level competition constraints.
- Developed transfer learning models for PM2.5 prediction, improving accuracy by 23%.
- Built a tropospheric O₃ prediction framework achieving ~89% accuracy using satellite data.
- Addressed data scarcity using domain adaptation and geospatial feature engineering.
- Delivered scalable ML pipelines for large-scale environmental monitoring applications.
Foundation
Education
Expertise
Technical Stack
Research & Leadership
Entropy-Guided Transfer Learning for Ground-Level PM2.5 Estimation from Satellite-Derived AOD over Indian Cities
IEEE INDICON 2025
Developed an entropy-guided transfer learning framework to address spatial data scarcity in urban air quality prediction, enabling knowledge transfer from data-rich to data-sparse regions.
Integrated satellite-derived Aerosol Optical Depth (AOD) with ground-level meteorological variables to construct a robust spatio-temporal prediction model.
Demonstrated a ~23% improvement over baseline models, highlighting enhanced generalization across heterogeneous urban environments.
Contributed a scalable, cost-effective methodology for air quality monitoring in regions with limited physical sensor infrastructure.
Uncovering Key Drivers of Tropospheric O₃ Prediction: A Remote Sensing-Based Machine Learning Framework
IEEE INDICON 2025
Designed a machine learning framework leveraging satellite remote sensing data to model tropospheric ozone (O₃) concentrations in the absence of dense ground monitoring systems.
Applied advanced feature engineering and model interpretation techniques to identify key environmental and atmospheric drivers influencing O₃ variability.
Achieved ~89% predictive accuracy, demonstrating strong model robustness and applicability in real-world environmental settings.
Established a reproducible data pipeline for large-scale atmospheric monitoring and decision support in developing regions.
Founding Co Organizer TEDxIIIT SriCity
TEDxIIIT SriCity is the first TEDx event in IIIT SriCity. Here I am Managing a team of 30 members from very diverse backgrounds. The Concerned teams are :
1. Event Management & Logistics
2. Speakers & Program Curation
3. Website & Tech Team
4. Marketing & Social Media
5. Video Production
6. The Design Team
7. Sponsorship & Finance Team
All these teams I am managing and with continuous discussion trying to make TEDxIIIT SriCity a success.
Find more at : https://www.ted.com/tedx/events/57968
Activity
Code Contributions
GitHub Activity
Connect
Let's build together
Pick the channel that fits best every link here stays in sync with my latest work.