Employee Since
May 7, 2015
I\’m a seasoned AI engineer specializing in computer vision. My expertise lies in crafting high-performing models for object detection, tracking, and image analysis, bringing intelligent vision to various domains. From optimizing real-time performance in autonomous vehicles to building scalable solutions for medical imaging, I thrive on tackling complex challenges through the power of computer vision. As a firm believer in MLOps practices, I ensure seamless model integration and continuous improvement. Driven by a desire to push the boundaries of this technology, I\’m eager to contribute my skills and collaborate on projects with real-world impact.
Skills
Expertise
- Natural Language Processing (NLP)
- Python
- Tensorflow
- Pytorch
- Llama 2 Language Model
Project Highlights
Conversational Chatbot using LLM and Semantic Chain (Llama 2,LangChain):
- Leveraging the power of the Llama 2 language model and the LangChain framework.
- I built a conversational chatbot capable of engaging in natural language interactions.
- Utilizing techniques like tokenization, intent recognition, and response generation.
- The chatbot learns from previous interactions to provide relevant and informative responses.
- The project demonstrates the potential of large language models for building engaging and intelligent chatbot applications.
Technologies:
- Natural Language Processing (NLP)
- Python,
- Pytorch,
- Tensorflow,
- Llama 2 Language Model,
LangChain Framework, - Pytorch
Intelligent Information Retrieval System:
- Designed and implemented an Intelligent Information Retrieval System using advanced NLP techniques to enhance document search and retrieval efficiency.
- The system incorporates semantic analysis, document summarization, and relevance scoring for precise and context-aware information retrieval.
Technologies:
- NLP
- Python
- Pytorch
- Tensorflow
- Word Embeddings
- Text Summarization
- NLTK
- BERT
- Semantic Search
Real-time AI-powered Theft Detection System using Computer Vision:
- Real-time video analysis with computer vision detects suspicious activities like unauthorized access and item removal.
- MLOps pipeline ensures continuous model training and deployment for optimal accuracy and adaptability.
- Customizable alerts and notifications empower security teams for swift intervention and loss prevention.
Technologies:
- NLP
- Python
- Pytorch
- Tensorflow
- transformer
- Opencv
- CNN
- LSTM
- Github Actions
- CI/CD Pipeline
- MLops
Continuously Evolving Traffic Sign Recognition System with Active Learning:
- Trains a computer vision model for recognizing traffic signs in real-time.
- Utilizes active learning to identify uncertain predictions and query human annotators for data labeling, improving model accuracy over time.
- MLOps pipeline automates data acquisition, active learning loop, and model deployment for dynamic adaptation to changing traffic environments.
Technologies:
- Python
- Pytorch
- Tensorflow
- OpenCV
- MLflow
- NVIDIA Triton
Machine Translation using Seq2Seq:
- Developed a machine translation system using a sequence-to-sequence (Seq2Seq) model to translate Urdu to English and vice versa.
- Implemented an encoder-decoder architecture with an attention mechanism, achieving accurate and fluent translations between the two languages.
Technologies:
- Python
- Pytorch
- Tensorflow
- NLTK
- Transformers
- Hugging face
- LSTM
- RNN
- Pytorch
Sentiments Analysis Using BERT:
- Implemented a sentiment analysis system using BERT (Bidirectional Encoder Representations from Transformers), a powerful pre-trained language model.
- Developed a pipeline that analyzes text inputs and determines the sentiment polarity (positive, negative, or neutral), aiding in understanding and classifying sentiment across different domains.
Technologies:
- Python
- Pytorch
- Tensorflow
- Hugging Face
- Tensorboard
- NLTK
- Spacy
Marathon Participant Retrieval:
- Designed and implemented an image clustering algorithm for identifying and grouping participants in a marathon based on the numbers on their shirts.
- Utilized clustering techniques to group similar images, enabling efficient retrieval and organization of participant data for event management purposes.
Technologies:
- Python
- Pytorch
- Tensorflow
- Pillow
- Face Recognition
- Agglomerative clustering
- Easyocr
- Gradio Tensorflow
Scalable Real-time Object Tracking with Efficient Inference
- Tracks objects in live video feeds using containerized models for efficient, distributed inference.
- CI/CD pipeline and MLflow integration ensure rapid deployment and performance monitoring.
- Enables high-demand applications like traffic analysis, crowd monitoring, and security surveillance.
Technologies:
- Python
- Pytorch
- Tensorflow
- Docker
- CI/CD Pipeline (Github Action)
- MLflow
- Model Pruning/Quantization