Data Scientist
Summary
Developed a quality monitor framework to evaluate LLM-generated responses by creating custom NER models for entity extraction, ranking responses based on contextual similarity, and implementing the ReAct agent, to compute precise scores on a 0-1 scale. Achieved 83% accuracy with a response time of less than 5 seconds.
Machine Learning Engineer
Summary
Developed and deployed a Liveness project, a facial recognition system to detect spoof faces, blurriness, masks, closed eyes, and multiple faces, achieving 85% accuracy with a 3-second API response time, and reducing annual expenses by $350,000. Implemented a document classification and field extraction pipeline for PAN, Passport, and Aadhar using CNN-based classification models, OCR, and NER techniques to classify documents and extract fields like ID numbers, and names, achieving 90%+ accuracy. Optimized resource management across multiple ML applications using multithreading, container tuning, and continuous monitoring (Prometheus, Grafana), achieving a 64% reduction in memory usage, a 47% decrease in core usage, and saving $20,000 monthly.
Machine Learning Engineer Intern
Summary
Enhanced Groww's search engine using transformer-based models and edit distance techniques to correct spelling errors and interpret user query contexts, improving search-to-search click suggestions by 15% with query inference time under 5ms. Worked on a large tabular dataset with millions of customer records by implementing resampling techniques for imbalanced data, applying feature engineering using PCA, and leveraging gradient boosting algorithms, achieving an area under the ROC curve of 0.96. Designed and deployed a scalable mandate form verification system using text and image classification techniques, detecting cropped images and validating signatures with 92% accuracy on real-world data.
→
Bachelor of Technology
Electronics and Telecommunication Engineering
Grade: 9.10/10
Courses
Statistics and Probability for Data Science
Calculus
Linear Algebra
Machine Learning
Natural Language Processing
Computer Vision
Awarded By
Omdena
Recognized as "Lead ML Engineer" out of 40+ ML Engineers at Omdena for developing a predictive analytics solution quantifying economic and environmental impact using tabular data.
Secured 3rd place out of 35+ teams in Syngenta's Machine Learning Hackathon for a computer vision task classifying seed germination stages based on root and stem presence.
Solved 700+ coding problems on algorithms and data structures, maintaining a consistent problem-solving streak for 1 year.
Awarded By
Toastmasters International
Received the Triple Crown Award and completed 4 levels of the Innovative Planning Path at Toastmasters International.
Machine Learning, Deep Learning, Natural Language Processing, Large Language Models, Computer Vision, Python, C++, PyTorch, Kubernetes, Deployment.
Project Management, Public Speaking, Mentorship, Teaching, Swimming.
Summary
Implemented a real-time object detection model on a GPU-enabled system for precise obstacle navigation, integrated enhanced 3D spatial awareness to detect walls and vehicles, and integrated OCR technique for real-time printed text recognition.
Summary
Developed a keyphrase extraction system for scientific literature to improve information retrieval, leveraging POS tagging, graph-based algorithms, and domain-specific filtering; ensuring high precision and semantic richness in extracted keyphrases.
Summary
Developed an LLM-powered system to enable natural language querying of tabular data using Text-to-SQL, and created the data analysis pipeline where users can enter their queries to perform tasks like visualization and interpretation of the data.
Summary
Developed a computer vision framework to distinguish plants from weeds in agricultural fields using object detection algorithms, achieving 96% accuracy. Processed datasets in XML format to structure the data for efficient model training and evaluation.