TensorsOfTheWallResearch: Where ‘what if’ turns into months of debugging and a publication that makes it all worth it.
Click on each card to view more details.
Research Intern|Full-Time|Kolkata, India
Achievements
- Led and implemented modified approach to person re-identification based on dynamic gait
- Achieved best accuracy of 91.13% on CASIA-B Gait Dataset (Python, Pandas, Numpy, Scikit-learn)
- Presented research paper at ETCCS 2020, indexed in Springer LNEE Series
Research Intern|Full-Time Co-Op|Bangalore, India
Received a full-time offer, which was deferred due to a hiring freeze during the COVID-19 pandemic
Achievements
- Adapted open-set classification to identify rare objects in vehicular datasets by merging classification models with meta-learning and few-shot algorithms, achieving 56% accuracy on the Open Images subset
- Developed novel algorithm to predict automobile part degradation and recommend preemptive actions
Teaching Assistant|Part-Time|Boston, MA
Achievements
- Assisted with an introductory course on cyber-physical and IoT systems
- Supported course logistics, conducted classes, graded assignments, managed lab sessions independently and contributed to developing new assignments
Visiting Researcher / Graduate Research Assistant|Full-Time|Boston, MA
Achievements
- Enhanced vision-language model performance for autonomous driving by 80% through creation of novel dataset and application of fine-tuning, RAG, and RAFT techniques
- Collaborated with Red Hat to optimize cloud-edge collaboration algorithms, developing a conditional routing module for dynamic offloading that reduced energy consumption
- Achieved state-of-the-art performance in CARLA-based navigation metrics, outperforming existing solutions by 35% while enhancing energy efficiency
- Designed a novel crowd navigation environment and introduced the Ecological Navigation Score metric to prioritize energy-efficient navigation in reinforcement learning tasks
- Developed visual servoing for multi-robot simulations using RGB and depth sensors, integrated human feedback, and evaluated reinforcement learning algorithms for person-following, significantly improving performance
Independent Researcher|Part-Time|Bangalore, India
Achievements
- Pursuing an original research idea in the intersection of AI and music, focusing on source separation. Compute generously sponsored by LossFunk Initialize.