Drashti Bhavsar

MSCS @NEU’25 || Ex- MLE @LOTI.AI || Prev. Researcher - GoogleExploreCSR @IIT-Roorkee, ISRO, IMD || PDEU CS’24 ||

Los Angeles, California, United States

About

With ongoing MS studies in Computer Science at Northeastern University, my focus lies in applying advanced machine learning techniques to solve complex problems. I developed a semantic search and ranking system for AI prompts, deepfake detection system, optimizing model efficiency by striking a balance between speed and accuracy. Skilled in applying cloud technologies like Microsoft Azure, AWS, and Amazon EC2, I am passionate about leveraging AI for impactful solutions. Let's connect and explore how to collaborate to amplify data literacy and innovation. Open to discussions on research, education, and all things data! 📈🔍

Experience

  • Head of Events at ACM at Northeastern University
    Jan 2026 - Present · 7 mos

  • Machine Learning Engineer at Loti
    Aug 2024 - Aug 2025 · 1 yr 1 mo

  • AI ML Researcher at Indian School of Business
    Feb 2025 - Jul 2025 · 6 mos

    Worked on large-scale Indian government patent data, performing end-to-end data analysis and NLP-based research from scratch. Cleaned, structured, and analyzed patent text to classify filings by gender representation and AI relevance. Applied BERT-based models and spaCy for text classification and entity recognition, and generated analytical insights to support research outcomes.

  • Mentee - ACM India Summer School at Indian Institute of Technology Gandhinagar
    Jun 2025 - Jun 2025 · 1 mo

    AI for Social Good

  • Junior Research Fellow at Indian Institute of Technology, Roorkee
    Jan 2024 - Jun 2024 · 6 mos

    • Selected as one of the top 50 students from a competitive pool of applicants for the Google-sponsored exploreCSR research mentorship program. • Led a project on autonomous driving maneuvers (lane changing, overtaking), enhancing accuracy by 25%. • Designed hierarchical Deep Reinforcement Learning models in CARLA simulator for autonomous vehicle navigation, training agent with depth cameras and obstacle detection modules. • Improved decision-making accuracy by 25% in dynamic environments through feature engineering and reward function optimization, solving complex overtaking and lane-changing problems. • Launched model achieving 94% success rate in simulated urban scenarios with real-time inference under 50ms latency.