class ShahirAnsari:
role = "ML Engineer & Computer Vision Specialist"
location = "Delhi, India 🇮🇳"
focus = [
"Object Detection", # YOLOv8 / Mask R-CNN
"OCR & Document AI", # Handwriting, Aadhaar extraction
"NLP & Chatbots", # HuggingFace Transformers
"ML Pipeline Design", # Training → Deployment
]
stack = ["YOLOv8", "PyTorch", "TensorFlow", "OpenCV",
"HuggingFace", "Keras", "scikit-learn", "Docker"]
building = "AI systems that see, read & understand the world"
fun_fact = "Trained a bot to play CS:GO — it outperforms me now 💀"
def say_hello(self):
return "Let's build something intelligent. 🤝" |
|
YOLOv8-powered document intelligence pipeline that detects and extracts key fields (name, DOB, UID) from Aadhaar card images with bounding-box precision.
|
End-to-end OCR system for reading handwritten text from images — applicable to number plate recognition, form digitisation, and field surveys.
|
|
A production-ready Telegram bot powered by HuggingFace Transformers — processes user messages through an NLP pipeline and responds in real time.
|
Trained a YOLO model on in-game screenshots to detect and target enemies in real time. A wild experiment where the bot eventually outperformed its creator.
|
|
Content-based and collaborative filtering recommender system built with pandas and scikit-learn. Suggests films based on genre, cast, and viewing history.
|
Implementation of Mask R-CNN (Matterport) on Keras + TensorFlow for pixel-level object detection and instance segmentation on custom datasets.
|
MULTIMODAL AI ████████████░░░░ Vision + Language models
DOCUMENT INTELLIGENCE ████████░░░░░░░░ Layout parsing, form extraction
ML PIPELINE DESIGN ██████░░░░░░░░░░ Training → serving → monitoring
EDGE DEPLOYMENT ████░░░░░░░░░░░░ Real-time detection on edge devices
// built with intent · powered by curiosity · driven by data