--> Rashmi Nagpal
Software Engineer

I am a Machine Learning Engineer at Cactus Research Labs. My primary research interests are in Fairness, Accountability, Transparency, Ethics (FATE), and Interpretability of NLP algorithms.

In July 2020, I graduated from the Plaksha & SCET UC Berkeley with a post-graduate degree in Artificial Intelligence. Previously, from 2018 to 2019, I worked at an Indian IT company in the Banking and Financial services sector, wherein my work revolved around the creation and deployment of deep learning models in production. In addition, I completed B.E in Computer Science from IIIT-Delhi in 2018. I deeply care about diversity and inclusion; subsequently, I volunteer my time teaching women from various socio-economic backgrounds in Go programming language as a global leader of Women Who Go, a non-profit organization.

Extracting Fairness Policies from Legal Documents
Rashmi Nagpal, Chetna Wadhwa, Mallika Gupta, Samiulla Shaikh, Sameep Mehta, Vikram Goyal
arXiv PDF
Distilling Knowledge From BERT Into Simpler Machine Learning Models
  • Investigated the potential of neural network distillation to interpret large black box models.
  • Proposed a simple and effective compression pipeline that transfers knowledge from BERT to simpler models using SVM, Decision Trees, Logistic Regression and achieved an accuracy of 87%.
Speech to Text Summarization
  • Developed abstractive summarization system using sequence-to-sequence models.
  • Achieved 70% score using BLEU(BiLingual Evaluation Understudy) to evaluate quality of generated text.
Malware Detection Engine
  • Developed a malware detection solution to classify portable executable files as benign or malicious.
  • Achieved an accuracy of 96% with a detection runtime of 3 seconds allowing for early remedial action using XGBoost model.
Machine Learning in Ruby
Intrigued by Artificial Intelligence but unsure how to get started using Ruby? In this talk, I will walk through a simple example to highlight the key concepts!
RubyConf 2021 - Denver, Colorado, USA
Software Is (Still) Eating The World
Technology advances at an ever-quickening pace. To benefit from these advances, system integrations need to be decoupled and polyglot. For an enterprise that wasn't born in the cloud and employs decades worth of legacy software, it's even more challenging to rethink the business as events or to shift from monolith architecture to event-driven microservices architecture. This talk will talk about the evolution and need of event-driven microservices and will demo an introductory example of building microservices in Go
4Developers Conference 2021 - Warsaw, Poland, EU
Learning Unsung Gotchas in Go
Go is one of the simplistic, neat, and effective languages, but it has few gotchas. These gotchas run the gamut from minor syntactic annoyances to psychological behavior. By sharing the invaluable references - this talk will cover them, which will conserve Go programmer's time and wasted effort. From beginner to expert level, everyone is warmly welcomed as the examples shared will cover a breadth of opportunities to learn.
GopherCon Russia, 2021 - Moscow, Russia
Fabricating Communities : One byte at a time
Can we build a welcoming, diverse and inclusive tech-community and learn the importance of it?I would love to share my experiences after getting involved in the Go community.
Go Conference 2021 - Tokyo, Japan
Deep Learning for Gophers
The software has not eaten the world yet, but infact has changed the way it was before. That software has given us, the human a new superpower which is the power of artificial neural networks. The goal of those networks is to help us answer the question : “Given X, predict Y with Z% accuracy”. This is where Deep Learning comes into picture. Let’s build a basic building block of deep learning : neural network.
FOSDEM, 2020 - Belgium, Brussels, EU