Mechanical Engineering

Designed and developed a 2-stage gearbox for a sugar industry from concept to delivery, performing all calculations, SolidWorks modeling, and GD&T implementation. Oversaw manufacturing coordination and delivered a fully functional unit, now successful in operation. The project ensured reliable power transmission with precise mechanical alignment and robust design.

Led the robotics team as Mechanical Lead for two consecutive years in the prestigious DD Robocon competition, designing and delivering two fully functional robots from concept to competition. Oversaw end-to-end mechanical development, including chassis design, mechanism innovation, manufacturing coordination, and rigorous testing.

Designed and manufactured a precision shaft with integrated key and coupling system, ensuring reliable torque transmission and alignment. Applied GD&T and tolerance stack-up analysis to achieve proper fit.

Designed and executed an experimental study on Powder Mixed EDM (PMEDM) of three super alloys (Inconel 718, Hastelloy, and Titanium), funded by ISRO. Focused on optimizing process parameters to enhance material removal rate, surface integrity, and surface quality for aerospace applications.

End-to-end design and fabrication of high-inertia flywheel assemblies for defence applications, focusing on rotational stability, material integrity, and precision balancing. Conducted stress and thermal analysis to ensure reliability under extreme operational conditions, while optimizing for weight and performance. Oversaw manufacturing, machining, and assembly processes.

End-to-end TuSimple lane detection pipeline using RESA, CLRNet-style models, and LSTR transformer. Includes data cleaning, training, decoder tuning, TuSimple metric evaluation, visual reporting, and JSON export. Designed for Kaggle T4x2 with a production-ready notebook.

Energy consumption prediction system using historical data, time-based features (lags, rolling stats), and ML models including tree-based algorithms and LSTM networks. Evaluated with regression metrics; final solution delivers accurate forecasts to support grid management and consumer optimization.

Real-time lane detection and departure warning system using a lightweight CNN trained on synthetic road images. Processes video at 50+ FPS, fits lane curves, and tracks vehicle position with hysteresis-based alerts. Fully self-contained and production-ready for ADAS on modest hardware.

IMDb sentiment analysis pipeline using TensorFlow Datasets, EDA (word counts, word clouds), and lemmatized text preprocessing. Trained Naive Bayes, Logistic Regression, and SVM on BoW and TF-IDF; Logistic Regression with TF-IDF achieved 89.1% accuracy. Incorporated GloVe embeddings to showcase advanced techniques. Delivered reproducible, well-documented code with error analysis and industry-grade practices.

Wine quality prediction using ML on combined red/white datasets. EDA, binary target (>6 = good), key features: alcohol & acidity. Trained Logistic Regression, Random Forest, SVM, XGBoost; Random Forest achieved 0.86 AUC. Offers insights to optimize wine production.

CycleGAN for unpaired image translation, implemented in a production-ready Jupyter Notebook. Includes generator/discriminator networks, training pipeline, visualization tools, and configurable hyperparameters. Demonstrates cycle-consistency with sample experiments and results.

This project explores the development of a haptic feedback device that enhances indoor cycling by simulating real-world road surfaces,smooth asphalt, rough gravel, and block pavement, through a stationary bike. The experience is paired with a virtual environment that visually matches the simulated terrain. A compelling fusion of physical sensation and digital immersion experiences.

Developed a low-latency, web-based control system for a robotic arm using WebSocket and Wi-Fi communication, enabling real-time, secure, and wireless operation. Designed a responsive graphical user interface that allows intuitive, precise manipulation of the robot’s movements while incorporating safety protocols to prevent operational hazards.

Developed a mobile robot system capable of precise object manipulation by synchronizing a 4-wheel omnidirectional base with a robotic arm for accurate end-effector positioning. Modeled the mobile base using a custom mobileRobot class in MATLAB, which converts wheel speeds into chassis twist and computes real-time pose updates via matrix exponential kinematics.