Engineering Autonomous Systems & Scalable Web Infrastructure
I am a final-year Engineering student at IIT Jodhpur, specializing in High-Performance Computing and Generative AI. My work bridges the gap between complex machine learning models and production-ready web applications.
I specialize in Agentic AI and Multimodal Data Pipelines. From engineering dual-engine architectures for developer tools to optimizing pricing algorithms for e-commerce, I build systems that are not just functional but intelligent and autonomous.
- Automating Developer Workflows: I build tools that write code. My project CodeBulb integrates Gemini 2.5 & DeepSeek V3 to create autonomous agents that self-correct syntax errors and scaffold projects via shell commands.
- Optimizing Business Metrics with AI: I don't just train models; I improve outcomes. I ranked in the Top 1% of the Amazon ML Challenge by reducing forecast error (SMAPE) by 20%, utilizing a 700+ feature pipeline with CLIP and ResNet50 embeddings.
- Scalable Full-Stack Engineering: I deliver speed and reliability. At SharpCareer Technologies, I reduced task-completion time by 45% for 500+ users by architecting modular React interfaces and robust REST APIs.
Python |
JavaScript |
TypeScript |
C++ |
PostgreSQL |
MySQL |
React.js |
Node.js |
Express.js |
PyTorch |
TensorFlow |
OpenCV |
Git |
Docker |
Linux/Bash |
Streamlit |
Postman |
Pandas |
- Stack: TypeScript, VS Code API, Gemini 2.5, DeepSeek V3.
- Core Tech: Engineered a dual-engine AI architecture capable of real-time context switching.
- Agentic Capabilities: Developed autonomous Node.js agents that execute shell commands to scaffold projects and auto-patch syntax errors without human intervention.
- Stack: Python, PyTorch, CLIP, ResNet50.
- Achievement: Ranked Top 1%. Reduced SMAPE from 66% to 35%.
- Deep Learning: Constructed a 723-feature pipeline fusing text embeddings (Sentence Transformer) and image features (CLIP/ResNet) to capture semantic and visual product data.
- Stack: JavaScript (ES6+), CSS3, jsPDF.
- Engineering: Interactive SPA visualizing O(n) shuffle algorithms using asynchronous DOM manipulation and regex-based log sanitization.
