Project Overview

  • This project was an applied engineering of the ESSP knowledge
  • The use case was to extract knowledge on alternate methods of polymer synthesis from literature
  • This was framed as an entity recognition and entity linking problem to help with both the component listing and the recipe composition
  • Just like other domain specific projects, this project required working with material scientists and chemists to build the initial framework
  • Then I had to build declarative and LSTM based ML systems to identify the required entities and their inter-links and communicate the findings with the domain experts
  • This information was later organized as an ontology with standard properties such as melting temperature, solvent etc. for each entity and labeled links between them

Skills

Applied Science

This project showed how scientific and technological processes can be applied to our everyday lives from the table we use to the car we drive. It was inspirational that a task very common in NLP, can be groundbreaking when applied to real life use cases

Working with material scientists and chemists

I picked up effective communication techniques that start with understanding usual terminology in one's expertise and trying to come up with analogies and bridges to share your expertise with them

Starting from nothing

This project had to start from zilch - no extracted data from polymer literature, no functional programming setup on the work stack the chemists use and no former POC guidance. This really helped me prove my handiwork and ability to bring structure and results to abstract problems