Ankana Sadhu

AI Researcher

Kolkata, West Bengal, India3 yrs 10 mos experience

Most Likely To SwitchHighly Stable

Key Highlights

Expert in deep learning architectures and custom algorithms.
Proficient in handling noisy data for predictive modeling.
Strong background in geospatial data processing and analysis.

Stackforce AI infers this person is a specialist in Agriculture technology with a focus on machine learning and data processing.

Contact

Skills

Core Skills

Machine LearningData Processing

Other Skills

C (Programming Language)C++Computer VisionDeep LearningGDALGenerative AIGoogle Earth EngineInternet of Things (IoT)Kalman filteringKerasPythonRasterioScikit-learnStatistical Data AnalysisTensorFlow

About

I specialize in research and applied areas—designing models, processing large-scale geospatial data, and building end-to-end pipelines for time-series and spatiotemporal prediction. I’m experienced in working with limited or noisy data and have designed custom algorithms to handle such constraints effectively. I have strong experience working with deep learning architectures, including CNNs, RNNs, LSTMs, and transformer-based models like GPT-2. I’ve also implemented classical ML algorithms like PCA, Naive Bayes, and Bayesian classifiers from scratch. My technical toolbox includes Python (with packages like TensorFlow, Keras, Scikit-learn, GDAL, Rasterio), MATLAB Simulink, and Google Earth Engine. I’ve applied these tools in projects involving satellite data analysis, generative adversarial networks for rainfall prediction, and crop yield forecasting using physics-informed ML. I’m also proficient in building loss functions tailored to imbalanced datasets, and in integrating domain knowledge into data-driven models. I scored 101 on the TOEFL and secured the 92nd percentile in GATE (ECE), which reflect my analytical and communication abilities in technical domains.

Experience

Office of industrial consultancy and sponsored research, iit madras

3 roles

Senior Research Fellow

Jan 2025 – Present · 1 yr 2 mos · Tirupati, Andhra Pradesh, India · On-site

Improving short to medium range extreme precipitation forecasts with Climate Networks and
hybrid physics-ML convective parameterization
Project Advisors: Dr. Arun K Tangirala (IIT Madras), Dr. Vishal Dixit(IIT Bombay) (Jan 2025-Present)

Senior Research Fellow

Promoted

Jan 2024 – Oct 2024 · 9 mos

Physics-based AI-ML Models for Predicting Crop Yield at different Space-Time scales.
Project Advisors: Dr. Arun K Tangirala (IIT Madras), Dr. Rojalin Tripathy (ISRO) & Dr. B.K. Bhattacharya
(ISRO)

Junior Research Fellow

Jan 2022 – Dec 2023 · 1 yr 11 mos

Physics-based AI-ML Models for Predicting Crop Yield at different Space-Time scales.
Project Advisors: Dr. Arun K Tangirala (IIT Madras), Dr. Rojalin Tripathy (ISRO) & Dr. B.K. Bhattacharya
(ISRO) (Jan 2022-Oct 2024)
⋆ Detailed study for selecting remote sensed data sources (satellite products such as MODIS, etc.).
⋆ Elaborate processing of the acquired data (images) using Google Earth Engine and Python packages: gdal, rasterio, etc to prepare dataset for the proposed model configuration.
⋆ Exhaustive research has been conducted to propose an algorithm, Multi-Stage Clustered (M-STAC) Prediction suitable to handle spatio-temporal variations in data, also with limited data samples (500-800 sample size).
⋆ The proposed algorithm, M-STAC Prediction provides a systematic method to tackle external factors to increase efficiency in the prediction from the traditional LSTM.
⋆ In the physics-informed ML model, the information derived using a crop simulation model [Agricultural Production Systems sIMulator (APSIM)] has been integrated within M-STAC.
⋆ Resources used: Google Earth Engine, Python Packages: tensorflow, keras, sklearn, gdal, rasterio,
tgan (Time-Generative Adversarial Networks)

Google Earth EnginePythonTensorFlowKerasScikit-learnGDAL+3

Indian institute of engineering science and technology (iiest), shibpur

Internship

Jun 2019 – Jul 2019 · 1 mo

My work primarily involved image processing using MATLAB. The model that was my object of study was a continuum robot, designed for the purpose of endoscopy. My job was to track the tip of the robot to obtain the final position of the continuum, for different angle of actuation of the smart actuator (dynamixel). The output and input data were fed into the Neural Network Tool of MATLAB to obtain the reverse kinematics of the system.