I deeply care about reproducible research and open science. Most of my code contributions can be found on my GitHub. A few glimpses from the Research/Software projects I've worked on, in the domains of Genomics, Machine Learning and Earth Observation, are:
Embedded Files
Monitoring Wildfire Danger using Deep Learning
deepGEFF forecasts global Wildfire danger by reproducing the capabilities of the Global ECMWF Fire Forecasting (GEFF) system using Deep Learning and developing further improvements in accuracy, geography and time scale through inclusion of additional variables and optimisation of it's algorithm, architecture & hyperparameters.
This open source project is developed in collaboration with ECMWF, Copernicus and WEkEO as part of ESoWC 2020.
deepGEFF Code on GitHub
deepGEFF Presentation Recording at ESoWC's AMA (July, 2020)
Automated Detection of Above Anvil Cirrus Plumes
Automated Detection of Above Anvil Cirrus Plumes
The above anvil cirrus plume (AACP) is a weather phenomenon that signifies an intense tropopause-penetrating updraft injecting cirrus clouds several kilometers into the stratosphere. Though AACPs can be identified by the human eye, no automated AACP detection methods currently exist. Lack of detection inhibits understanding of where and how often AACPs occur, and influence on stratospheric air composition.
This study demonstrates automated emergent AACP identification based on a U-Net model using only satellite infrared and visible reflected sunlight imagery. Model performance is quantitatively benchmarked with human AACP identifications and qualitatively assessed through animated detections from GOES-16 1-minute temporal resolution imagery.
Presentation Recording at EEML
Poster presented at NASA Langley Research Center (Nov, 2018)
Poster presented at American Meteorological Society 100th Meeting
Paper at 19th Conference on Artificial Intelligence for Environmental Science
Cancer Immunologic Data Commons (CIDC)
Cancer Immunologic Data Commons (CIDC)
The CIMAC-CIDC Network is an NCI Cancer Moonshot initiative that provides cutting-edge technology and expertise in genomic, proteomic, & functional molecular analysis to enhance clinical trials in cancer immune therapies. CIDC serves as the bioinformatics needs of the network, optimizing of data collection methodologies suitable for immune-related biomarkers, and building a biomarker database for the immuno-oncology community. I've learnt a great deal about Product and Design especially during the course of the CIDC Design Sprint .
CIMAC-CIDC Network Website
Code on GitHub: Schema, Command-Line Interface, API
Publications:
Multi-omics analysis reveals immune features associated with immunotherapy benefit in squamous cell lung cancer patients from Phase III Lung-MAP S1400I trial
Posters at the AACR Annual Meeting 2022/23:
Immune gene expression signatures associated clinical benefit from nivolumab and ipilimumab for previously treated patients with stage IV squamous cell lung cancer: An immune biomarker analysis of phase III SWOG LungMAP S1400I trial: view here
Infiltration and spatial distribution of immune cells are associated clinical benefit from Nivolumab and Ipilimumab for Previously Treated Patients with Stage IV Squamous Cell Lung Cancer: an immune biomarker analysis of Phase III SWOG LungMAP S1400I trial: view here
SMART: Space Metagenomics for Planetary Protection
SMART: Space Metagenomics for Planetary Protection
I developed a metagenomic analysis pipeline using open source bioinformatics tools to classify and characterize surviving microorganisms in the NASA JPL Spacecraft Assembly Facility, NASA Extreme Environment Mission Operations (NEEMO) facility, and the International Space Station. The pipeline performs trimming and quality filtering, assembly, annotation, and profiling on the sequenced samples collected to find their relative abundance and diversity.
Three bacterial phyla: Actinobacteria, Firmicutes and Proteobacteria dominated the ISS and two Earth-based cleanroom studies, but varied in their abundances. The dominant surviving species are observed to be tolerant to cleanroom/laboratory sterilization methods and are capable of survival in extreme and stressful environments.
Presentation at Caltech/NASA JPL (July, 2017)
Roshni Biswas, Nitin K Singh, Kasthuri J Venkateswaran, A New Metagenomics Pipeline Developed from Open-Source Computational Tools for Mars and ISS Missions, NASA JPL
Climate Change Impacts on Land Cover Change
Climate Change Impacts on Land Cover Change
The NASA Carnegie Ames Stanford Approach (CASA) is a global simulation model that combines multi-year satellite, climate, and other land surface databases to estimate biosphere-atmosphere exchange of energy, water, and trace gases from plants and soils.
In this project, I developed machine learning algorithms using the geo-spatial data used by the CASA Express model specifically to tackle climate change problems. These include mapping land cover change and performing temporal analysis of Greenhouse Gas emission at various geolocations using MODIS satellite data.
Presentation at NASA Ames Research Center (May, 2019)
Roshni Biswas, Andrew Li, Christopher Potter, Programming and Analysis of Land Cover Change, NASA Ames Research Center
Wikilimo: Climate Informatics for Precision Agriculture
Wikilimo: Climate Informatics for Precision Agriculture
Wikilimo is an online-offline platform for hyperlocal weather and agri-insights that analyses citizen science data and satellite imagery to provide actionable insights on agriculture, climate and finance to farmers in remote communities with limited connectivity.
The Wikilimo app was conceptualized as part of UNDP's Unite Ideas Summer of Solutions Contest and was adjudged the winning solution at Climate Week NYC 2019.
Perfect-Pitch: Objective Assessment of Pitch Accuracy
Perfect-Pitch: Objective Assessment of Pitch Accuracy
A sawtooth waveform inspired pitch estimator to determine and plot fundamental frequency levels from audio signals, this is an approach for objectively assessing accuracy of pitch in vocal monophonic music using signal processing techniques.
Such an evaluation system with quantitative analysis coupled with visual representation will greatly aid the training process of singers.
Report for NITK Surathkal SIP 2017
Paper at 6th International Conference on Advanced Computing, Networking, and Informatics (ICACNI)
Report abuse