My introduction to computer programming was serendipitous. Practicing it influenced my perception, analysis, and matured me into a self-learner; this blog is a culmination of that. I like programming in C++ because it offers abstractions to write readable code and room to build complexities. After that, I love the R language because of its open-source nature and the beautiful RStudio. It gives me the power to chase my Scientific curiosities. I also work with Processing occasionally on personal projects. At work, I use MATLAB. Just like in C++, I enjoy the problem of vectorizing operations in MATLAB.
This blog is a space for my scientific investigations, and I aim to present the findings in a way that allows readers to follow the chain-of-thoughts as they lead me to conclusions.
Topics I am currently pursuing in Complex systems
Understanding the Mathematical structures that govern overall behavior: Many Mathematical models for chronic illnesses are steady-state ODE models. Depending on the question, there could be several interacting agents, resulting in long simulations. Characterizing the undulations in the parameter space-regions of low and high sensitivity then becomes very expensive. A simple solution is to solve the system of ODEs for the steady-states, but the nonlinearity in the interactions prohibits one from doing a simple matrix inversion. So, a numerical estimation is the next best idea. There is an R library called 'rootSolve' to solve the problem. But there can be several steady-states in the phase space, and my attempts at solving a couple of non-linear systems failed to reach the right state from given initial conditions. This led me to deliberate on how to guide the numerical solvers to reach the right steady-states. This set me on a journey to design a process for uncovering Mathematical properties in these large interaction systems.
The connection between scales: Knowing how different scales are connected helps move between them from a known starting place. For, e.g., In clinical trials, phenotypic measurements and adverse reactions are used to categorize the disease severity and effectiveness of treatment. This informs clinicians to either continue the treatment or switch to a different one. Knowing how information is connected between scales, we could use clinical measurements to decipher cellular activities, adjust the dosage, give supplementary therapies to cause the desired reactions, or improve the drug's design. A similar initiative is already underway: pathways leading to adverse events. I briefly worked on designing clinical measurements from cellular interactions. Thorough knowledge in this improves the applicability of computational models to identify patients' sub-classes that will benefit most from the intervention. A similar work I did in this area was to design a scalar objective function for a multi-objective problem. This task required me to understand the information individually and collectively. This inspired me to delve deep into understanding the flow of information between scales.
Harvesting Energy using Vibration Absorber (Master's Thesis): An analytical study of coupled linear Electro-Mechanical system to understand the energy harvested from stochastic motion using an Energy Harvesting Dynamic Vibration Absorber (EHDVA). The outcomes are power output at the piezoelectric and the effects of mechanical, electrical, and material parameters on the power. This work is published on IFAC-PapersOnLine.
Network Meta-analysis: A cross-trial comparison of Immune Checkpoint Inhibitor Therapies (CIT) using clinical outcomes. My team and I used the univariate Bayesian method of analyzing the trials' network, inconsistencies, and SUCRA scores to rank the therapies. I gave a talk on this at the International Conference on Immunology 2018, held at SRM University.
Vantage Rheumatoid Arthritis (RA) QSP model: I led the Vantage RA QSP model is a multi-scale, steady-state dynamical system with interactions between immune cells and local cells at bone joints. My team and I developed the DAS28-CRP clinical score from the lower scale cellular interactions—link to a poster presented at PAGE 2019, Stockholm, Sweden.
Vantage VQM tools: This is a free and open-source tool in MATLAB to develop and apply QSP models. The tool is available for use in GUI, CLI, and embodies the practices for code readability, easy-to-use documentation. The code is hosted on Gitlab, and a poster on the tool is presented at PAGE 2019, Stockholm, Sweden.
Immuno-Oncology modeling: Worked with a multi-scale dynamical model capturing interactions between Immune cells and Cancer for studying the efficacy of novel drugs and their combinations. I designed a model calibration program for the multi-objective system using the Genetic Algorithm in MATLAB. Taking advantage of vectorizing operations, I reduced calibration time from several hours of run time to slightly over 1hr. This work is presented at ACoP11, 2020, THU-067.
Optimal nesting: Implemented a dynamic programming approach for nesting of rectangular objects in C++11.
I developed an interest and honed teaching skills through opportunities that came my way.
- Mentored newly recruited colleagues in their training periods and transferred knowledge on wide-ranging topics like Code vectorization, concurrency, and parallelism in C++, Multi-scale modeling, Importance of functional forms in Computational models
- Introduced Arduino programming with a hands-on exercise to school students at Swami Dayananda Higher Secondary School, Manjakkudi, Tamil Nadu. This was sponsored by Vantage Research and Atal Tinkering Labs (ATL) program.
- Gave lectures and did paper review presentations to my colleagues
- Worked as a teaching assistant at IIT Madras
Sport programming: Currently ranked under 350 on HackerEarth (username: Protagonist). Qualified to Bronze leagues in bot programming contests Code of fire & Ice, Code a la mode community contests on CodinGame (username: madpro3), Google CodeJam 2019.
Received a scholarship to attend the Robo-revolution workshop: Together with an interdisciplinary team, I built a light painting robot using Arduino. My key contribution was developing a program for circumventing tedious custom design of track for new images. The workshop was conducted in collaboration between Swissnex India and Workbench projects, Bangalore, India, 2015.
No comments:
New comments are not allowed.