• Developed a data pipeline for 8+ years of Facebook campaign data on Azure using Azure Data Factory, Azure Data Lake
Storage, and Azure Databricks to process 180GB of data from the Facebook Marketing API.
• Enhanced financial forecasting and risk assessment with advanced models including ARIMA, Prophet, Logistic Regression, and
Bayesian Structural Time Series using Python, optimizing Facebook ad targeting by 12% through audience segmentation.
• Integrated the Facebook API, utilized web scraping, and deployed GPT-4 API with AWS Lambda to generate the top 20 targeted
suggestion tags, leveraging data from the Data Lake.

• Deployed a GPT-4o pipeline, utilizing prompt engineering, GPT-4o Assistants, web scraping, and vector embeddings to create an RAG framework for DIB tag predictions.
• Deployed 7 advanced simulation algorithms using LSTM-based neural networks and Natural Language Processing (NLP) on AWS Lambda, achieving over 95% accuracy in predicting ad campaign outcomes such as reach and revenue.

• Simulated a bio-inspired, pneumatically-actuated 7-chamber soft snake robot using MuJoCo and SoFA Framework,
modeling robotic mobility and structural integrity.
• Engineered a computer vision tracking system using YOLOv8 to detect structural changes in real-time, employing pattern
recognition of 4 etched markings to identify chamber bursts, improving detection accuracy by 95%.
• Integrated a reinforcement learning model on AWS Sagemaker using PyTorch to formulate the top 3 adaptive strategies for the
robot using SoFA, enabling autonomous recovery from structural damages and achieving a 70% improvement in recovery time.
• Simulated bursts and punctures in SoFA Simulation, advancing the robot’s resilience through detailed modeling and stress
testing with reinforcement learning and real-time computer vision.

• Spearheaded the development of "Angelia" IDE extension to log student keystrokes in Visual Studio Code and PyCharm using
JavaScript and Java.
• Performed data analysis (300+ key logs, 10,000+ keystrokes) and employed supervised and unsupervised machine learning
techniques using Scikit-Learn to predict and assist students in overcoming coding challenges.
• Employed GPT2 and BERT-based LLMs for "Examinator" system to identify cheating instances in student exam submissions,
reducing search space from 234,000+ to 250 and cutting the time to detect cheating instances by 85%

Manager: Ms. Nishita Shashtri

Work:

• Automated the cell biology protocol onto a liquid-handling robotic platform, essentially helping to design, create, test, and integrate components and processes to allow automation of BreakSight’s assay.

• Revamped hardware and software components of the protocol, to optimize operational efficiency.

• Coded for thermal cameras, harnessing thermal imaging data for analysis and managing heating/cooling systems. Used Raspberry Pi to interface with diverse hardware components in Python 3 to automate the process.

• Constructed a 4 part design for thermal cameras, proficiently programmed cameras to regulate heating/cooling systems.

• Devised negative pressure system to extract liquids from cell samples, and successfully assembled and programmed liquid cooling systems for precise temperature control of cell samples.

​

Skills: 3D CAD, Python, API, Robotics, Mechatronics

Course: CS 3600 (Introduction to Artificial Intelligence)

Professor: Dr. Thad Starner

Work:

•  Conducted end-of-class 15 minute specialized sessions, providing practical insights into subject matter during class, fostering engagement and understanding. Resolved student inquiries, guided assignments and quizzes, and maintained vigilance over the Ed platform to promptly address and resolve student concerns.

• Safeguarded academic integrity by performing comprehensive plagiarism checks on 900+ submitted assignments, codes, and
  exams.

Advisor: Dr. Ye Zhao; Team Lead: Ted Tyler

• Led a team of 18 people to design and build a sensor-integrated (5 sensors) robotic foot that classifies the terrain that the robot is walking on thereby optimizing the robot’s gait, and improving stability.

• Finite Element Analysis of weak point on foot that fails due to fatigue.

• Trained multiple Machine Learning and neural network models for classification of terrain (99% accuracy).

• Generated CAD models and fabricated the different components, designed a modular terrain bed for testing.

• Spearheaded the production of a new, in-house & more robust tactile sensor.

• Signal processing (FFT, Kalman Filters, etc) for IMU Slipping Detection, and Capacitive and Acoustic Sessor data.

•  Designed a Gelsight Sensor for Pressure Sensor detection based on an MIT sensor.

Skills: Eagle PCB, Deep Learning, Fabrication and Analysis, Signal Processing.

Advisor: Dr. Aaron Young; Team Lead: Jennifer Leestma

• The overarching goal of the project is to develop an intelligent robotic hip exoskeleton to augment biomechanical stability during walking.

• Designing and testing a new 2-degree-of-freedom hip exoskeleton that will be used to study balance augmentation.

Roles: Preparing the exoskeleton hardware and software for real-time use, developing control algorithms and machine learning models for the exoskeleton, analysis of biomechanical data, assist in pilot testing and formal human subjects experiments.

Skills: Eagle PCB, Deep Learning, Robot Operating System (ROS),            SOLIDWORKS

Manager: Mr. Atul Mishra

• Innovated a self-service reporting tool in APG(Aladdin Product Group), team of 4, that reduces 2 weeks of manual work per report.

• Integrated authentication functionality to Balancer Blockchain protocol.

• Resolved 20+ support tickets that required database modifications and software reconfigurations.

​​

Skills: Python, Java, Angular, Solidity

Mentor: Dr. Dhiraj Mahajan

• Created a Hydrogen-EV hybrid crop-hopping drone with a modular structure for weeding, surveillance, and pesticide spraying.

• Managed a team of 5.

• Designing a modular drone from scratch.

• 3D modeling and MATLAB simulation for the whole modular structure.

• Machine Learning and computer vision for crop disease surveillance and prediction, model created using CNN and Transfer Learning.

​​

Skills: CAD, Transfer Learning (CNN), MATLAB, Simulink

President: Vishu Chauhan

• Managing a team of students for different projects like Project Meera, Project Naritva, etc.

• Leading the sub-project Shreeraksha, women security device under Project Naritva.

• Technical Mentor for fresher and sophomore students in Enactus Club.

​

An online Photo-blog platform modelled from Humans of New York for sharing stories of Thapar University Residents.

As of July, 2021:
- 2400+ followers
- 22,000+ total impressions
- 6000+ reach to audience
- 1000+ interactions per week