Charmi Shah

Experience

• 

  Analysed.in

  Oct 2020 - Jan 2021

  Machine Learning Intern

  - Worked on designing and implementing a feature to display the interview progress of a candidate to the recruiter using PHP.- Designed and implemented a chatbot and integrated it with LinkedIn using BeautifulSoup framework.- Extended a face-recognition model for monitoring purposes in the video-calling feature for the website.

• 

  SVNIT, SURAT

  Jun 2021 - Apr 2022

  Head of Content Writing

  Led a team of 14 content writers and created articles, speeches, and stories for the Academics Affairs Council of SVNIT for the year 2021-2022. This role not only enabled me to develop my management skills but also encouraged me to venture into uncharted creative territories.

• 

  Ugam

  Jun 2021 - Aug 2021

  Data Analyst

  - Built a semantic search engine using graph databases for retail applications for contextual AI. - Created a knowledge graph on retail data combining different categories of products and materials used for products.- Implemented and built graph database from scratch for the same which reduced result latency by 30%.- Explored and researched ways to centralise the data to make it available in multiple languages reducing cost and storage.

• 

  Amazon

  Jul 2022 - Mar 2023

  SDE

  - Developed a console to automate the onboarding process of eight critical internal services (CloudAuth, Switch etc) with a capacity of handling over a million requests per second.- Designed and developed a monitoring system that enabled customers to dynamically access their resource metrics while ensuring the principle of least privilege. - Worked on creating resources with CDK with cross-account AWS resource communication.- Identified and wrote a few critical test cases, enhancing overall coverage. Show less

• 

  Merkle

  Mar 2023 - now

  Senior Analyst

  - Leveraged GenAI for extracting insights from data of our clients with conversations among individuals who had encountered fraudulent ads, allowing us to proactively detect and mitigate such deceptive practices.- Enhanced the development of models powering the portal's analytical capabilities by leveraging BERT for sentiment analysis, achieving an 80% accuracy rate. Additionally, employed spaCy for entity recognition, attaining an 86% accuracy rate, and enabled topic modeling with a 0.01 coherence score.- Examined and enhanced processing efficiency by optimizing performance, resulting in a 40x reduction in processing time by implementing multi-threading techniques - Initiated and implemented comprehensive documentation for a project responsible for computing and maintaining all model training, interpretation, and examination of the data, uploaded by over 2,000 users. Show less