class ManasDeshpande:
def __init__(self):
self.name = "Manas Deshpande"
self.role = "B.E. Electronics & Computer Science (2023–2027)"
self.college = "Vidyalankar Institute of Technology, Mumbai"
self.focus = ["Machine Learning", "Computer Vision", "RAG Systems", "Full-Stack AI"]
self.stack = ["Python", "TensorFlow", "YOLOv8", "LangChain", "FastAPI", "React"]
self.currently = "Building end-to-end AI pipelines & hunting for AI/ML internships"
self.fun_fact = "I trained a UNet to find cracks in bridges before I could drive one 🌉"
def say_hi(self):
print("Always shipping, always learning.")
|
Real-time multi-class object detection over live video streams using YOLOv8, with a Next.js dashboard for monitoring, alert notifications, and detection history.
|
Automated crack localization in concrete and pavement using UNet and CNN, trained on real-world infrastructure images with data augmentation for robust generalization.
|
|
RAG-based AI assessment engine scoring prompts across 6 weighted criteria using FAISS, sentence-transformers, and Gemini 1.5 Flash — with exponential-backoff retry logic and file-based caching.
|
ML web app predicting prices across 25+ Indian cities using an ensemble of Linear Regression, Random Forest and Gradient Boosting — served via Flask REST API with glassmorphism UI.
|
- 🔭 Building production-grade AI systems with FastAPI + React
- 🌱 Deepening expertise in Agentic AI and LLM fine-tuning
- 🧪 Expanding my RAG evaluation framework (PELS) for broader domains
- 📖 Exploring PyTorch and MLOps pipelines
- 💼 Open to AI/ML internship opportunities across India (remote-friendly)
| Platform | Link |
|---|---|
| 🌐 Portfolio | manasd2011.github.io/portfolio |
| manas-deshpande21 | |
| manasdeshpande51@gmail.com | |
| 💻 GitHub | ManasD2011 |