KESHAV THAKUR

• Partnered with researchers and PhD students at IIT Bombay to investigate the theoretical and practical underpinnings of convolutional and recurrent neural networks in classification and sequence modeling.
• Explored the internal mechanics of CNNs for spatial pattern recognition and RNNs for time-dependent data analysis, including LSTM and GRU architectures.
• Applied advanced mathematical concepts (e.g., matrix calculus, gradient descent, eigen decomposition) to improve model training and interpretability.
• Built, tuned, and tested neural network models in Python using Jupyter Notebook, analyzing accuracy, loss, and generalization behavior across experiments.

• Analyzed problems and worked with teams to develop solutions.
• Participated in workshops and presentations related to projects to gain knowledge.
• Explored new technologies and approaches to streamline processes such as MsQuic protocol.
• Worked with different teams on projects, and encountered different personality types.
• QUIC protocol project in conjunction with hardware device, Data Labeling project where I worked with highly confidential processes in the Intel fabrication plant.

• Participated in workshops and presentations related to projects to gain knowledge.
• Explored new technologies and approaches to streamline processes.
• Learned basics of convolutional neural networks
• Learned basics of self-driving ML software, object detection

Taught underprivileged children in a village in India the basics of building a computer, navigating the Internet, and basic programming including Scratch and Java.

Aided and taught basic math concepts to children with autism (60 hours)