Dhaval Parmar

June 2021 - Present

ConcertAI, Bangalore

• Experimented with various LLMs and techniques to do the feasibility studies for different product use cases
• Built ML models to extract different entities from notes and worked to improve existing model performance
• Implemented Data Version Control (DVC) for code bases to improve tracking
• Maintained and improved tools important for data validation work

April 2020 - July 2020

Philips India Ltd, Bangalore

• Worked on automatic landmarks detection corresponding to corpus callosum in brain MRI images using reinforcement learning
• Used Q-learning based Deep Q-Network (DQN) and its variants Double DQN, Dueling DQN and SARSA algorithm

Dec 2023

65th American Society of Hematology (ASH) Annual Meeting

Built XGBoost based machine learning models for patients with Higher-risk myelodysplastic syndrome (HR-MDS), towards predicting time-to first line systemic therapy which get accepted for poster presentation.

Nov 2023

International Society for Pharmacoeconomics and Outcomes Research (ISPOR) Europe Conference

The poster selected for presentation describes XGBoost based time-to-event modeling approach to predict overall survival as event for patients with MDS using clinical data.

June 2023

American Society of Clinical Oncology (ASCO) 2023

The poster represents benefit of NLP based models developed by us to extract information from unstructured Electronic Health Records to enrich structured data withhigh accuracy which helps in many downstream tasks.

Oct 2022

The 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2022)

Idea is to learn multiple policies sequentially on top of the previously learnt policies in option framework setting. As part of dissertation project, contributed by considering Duckietown environment and goal based navigation task.

Nov 2020

9th IEEE International Conference on Cloud Computing in Emerging Markets (CCEM)

Paper presents my work during internship to compare Reinforcement Learning algorithms and Deep Learning based approaches for landmark detection tasks. The RL model is also able to learn locations of multiple landmarks simultaneously.

2019 - 2021

Indian Institute of Science, Bangalore

• Learnt about machine learning, deep learning and reinforcement learning techniques in different domains like NLP, CV and Speech
• Studied fundamental concepts behind AI like linear algebra, probability theory, optimization, Data structures and algorithms

2013 - 2017

Faculty of Technology and Engineering, Baroda

• Learnt about microcontrollers, microprocessors architecture and related assembly programming langauge
• Implemented algorithms to solve polynomial equations of real world problems in power system

Coursera

• It includes four courses covering basic concepts about data centric AI to machine learning models in production
• Learnt about architecture and hyperparameter search, serving infrastructure, importance of testing and monitoring

Udemy

• Through hands on examples, learnt about creating ML pipeline, packaging it and deploying it using REST API
• It also explains about CI/CD cycle of pipeline and deploying pipeline using container on AWS

Udemy

• Learnt about implementing different tests like unit, differential, integration and testing in production using shadow mode
• Come to know about monitoring metrics with Prometheus, Grafana and logs with Kibana, Elastic search