01
Percepta
A clip-on camera for glasses that reads the street ahead and speaks hazards aloud, built to help visually impaired people move through it.
AI workflows, Code, Design.AI Engineer.
I create engaging, 
unique experiences that matter.
Ideas get
better together.
Selected Work .Design
Then it has to work.
I take machine learning fromprototype to product.
def forward(self, x):
x=self.pool(F.relu(self.conv1(x)))
x=self.pool(F.relu(self.conv2(x)))
x=x.view(-1, 16 * 4 * 4)
x=F.relu(self.fc1(x))
x=F.relu(self.fc2(x))
x= self.fc3(x)
return x
0+0+ - Hackathons Won
0+0+ - Projects
0+0+ - Repositories
Machine Learning
Agentic Development
Searched repository for pytorch training pipeline
The code fixes gradient accumulation by clearing .grad after each optimizer step. I’ll verify that placement matches nearby examples and doesn’t change first-step behavior.
Used Figma Integration
I’ve verified the gradient-reset change. I’ll reflect it in the Figma slide with a focused before-and-after diff, a one-line explanation of its impact, and the test status.
Loaded a tool, ran a command
Edited a file, ran a command
Edited 1 file+138-67
Review GitHub and Figma, then use main.ipynb to improve the PyTorch training pipeline. Run tests, then summarize changes in README.md.
