Experience • Nirav Shenoy

Relational ML Lab - CISPA

Worked on building deep learning algorithms to perform efficient model pruning for vision models.
My thesis was on the topic of Efficient Pruning: Combining Continuous Sparsification with Dynamic Sparse Training.
Achieved 95%+ sparsity in ResNet models on ImageNet-1k and CIFAR10/100 with comparable accuracy to dense models, with an 80% inference runtime reduction.

QUASIM was a project based on applying graph-based quantum machine learning models for prediction and optimization tasks in industrial laser cutting simulations.
The results of our experiments were submitted to the Quantum Machine Intelligence 2025 and presented at TddT Berlin 2024.
Developed EAVE dashboard as part of ESCADE with FastAPI, React, Next.js and Docker for energy-efficient AI in data centers.
Presented to stakeholders at Hannover Messe ‘25
Our research was submitted to CAiSE 2025 under the Research Projects Exhibition track

Startup where I was responsible for building a Generative AI chatbot that assists MBA candidates in India with their B-school applications
Implemented an RAG system that could help provide candidates with up to date and detailed information about the colleges that they were interested in applying to.
Vector database includes up-to-date information of the top 200 B-schools in India

Designed and implemented scalable cloud solutions for over 3600 customers using Java, Springboot, and HANA.
Led the SuccessFactors Job Agency team in building a pilot end-to-end microservice using Docker and Kubernetes.
Modularized cloud application handling 25+ million requests per day into smaller services.
Reduced customer defects and security vulnerabilities for the Agency module by 50% over 6 months.

Created a Speech-to-Text API PoC with HMMs and a Random Forests model to enhance mobile app usage at airports.