Prajjwal Yash

Scientist at Indian Space Research organization knowledgeable in Python programming and has worked on Machine Learning challenges for two years. Determined and well-rounded individual with 2 years working as Scientist within Space industry.

Football analytics using Python (MadAboutSports)

1. Master's thesis - Quantum critical aspects of the Superconductor-Insulator Phase transition in disordered Superconductors

The effect of increase in disorder strength for Hubbard model at half filling with the repulsive potential acting as the disorder strength depending on a site being singly- or doubly- occupied is observed in the form of a metal-insulator transition(MIT) due to localization of states . The critical aspects of this quantum phase transition are investigated and important inferences of the same are obtained. In order to study the interplay of superconductivity and localization, an attractive potential (BCS) is further introduced and the effect of disorder and attractive interaction is investigated.

2. Power System health monitoring using Machine Learning

Time-series obtained from spacecraft (telemetry) is used to monitor the health of the satellite in space. An ML architecture has been developed which uses multiple algorithms - from unsupervised time-series clustering to auto-regressive time-series forecasting models - in order to detect the present state of the power subsystem of a spacecraft and detect possible off-nominal behavior. With the domain expertise available for the system in question, integrating the same into the ML architecture has shown better performance and proven to be a more attractive preposition than an off-the-shelf AI/ML tool.

3. Computer Vision project - Detection of cracks in space grade solar cells

A novel method of obtaining explainable ensembled decision-making from a Convolutional Neural networks (CNNs) is developed. Explainability is an extremely important criteria when working on space quality AI solutions as the criticality is high. End-to-end automation has been performed and the solution is deployed on a high-performance central cluster.

4. Training Project for ISRO - Estimation of solar irradiance at Martian surface for prospective lander/rover

A quantitative estimate of available solar power on Martian surface throughout a Martian Year (~687 Earth days) is obtained. The interaction of sunlight with atmospheric dust, water ice clouds and gaseous content of the atmosphere is modeled. The quantitative analysis obtained brings to light the difficulty of navigating exploration of Mars during the period from the perspective of power generation for the life of the mission.

1. Master's thesis - Quantum critical aspects of the Superconductor-Insulator Phase transition in disordered Superconductors

The effect of increase in disorder strength for Hubbard model at half filling with the repulsive potential acting as the disorder strength depending on a site being singly- or doubly- occupied is observed in the form of a metal-insulator transition(MIT) due to localization of states . The critical aspects of this quantum phase transition are investigated and important inferences of the same are obtained. In order to study the interplay of superconductivity and localization, an attractive potential (BCS) is further introduced and the effect of disorder and attractive interaction is investigated.

2. Power System health monitoring using Machine Learning

Time-series obtained from spacecraft (telemetry) is used to monitor the health of the satellite in space. An ML architecture has been developed which uses multiple algorithms - from unsupervised time-series clustering to auto-regressive time-series forecasting models - in order to detect the present state of the power subsystem of a spacecraft and detect possible off-nominal behavior. With the domain expertise available for the system in question, integrating the same into the ML architecture has shown better performance and proven to be a more attractive preposition than an off-the-shelf AI/ML tool.

3. Computer Vision project - Detection of cracks in space grade solar cells

A novel method of obtaining explainable ensembled decision-making from a Convolutional Neural networks (CNNs) is developed. Explainability is an extremely important criteria when working on space quality AI solutions as the criticality is high. End-to-end automation has been performed and the solution is deployed on a high-performance central cluster.

4. Training Project for ISRO - Estimation of solar irradiance at Martian surface for prospective lander/rover

A quantitative estimate of available solar power on Martian surface throughout a Martian Year (~687 Earth days) is obtained. The interaction of sunlight with atmospheric dust, water ice clouds and gaseous content of the atmosphere is modeled. The quantitative analysis obtained brings to light the difficulty of navigating exploration of Mars during the period from the perspective of power generation for the life of the mission.