Subrakanth Sutar

Predictive Rounding Model.

Directed the development of a machine learning model to address class imbalance, enhancing prediction accuracy for the minority class in large datasets.
Applied techniques like SMOTE and Random Under Sampling, and adjusted class weight parameters to optimize F1-score and other key metrics.

Predictive Rounding Model (Fairness Analysis)

Led the integration of a fairness analysis pipeline within ML flow using Aequitas, focusing on ethical AI practices.
Analyzed and corrected disparities related to gender, race, and region, ensuring fair and transparent outcomes in machine learning models.

Technical Skills
Languages: Python
Tools/Platforms: DeBERTa, RoBERTa, XLNet, DistilBERT, Aequitas, IBM AI Fairness 360
Techniques: Machine Learning, Deep Learning, Sentiment Analysis, Bias Mitigation, Fairness Analysis

NDx (Sentiment Analysis Scoring)

Fine-tuned advanced NLP models (RoBerta-Large, XLNet-Large, DistilBert-Base) for sentiment analysis, achieving a 97% accuracy rate in both regression and classification tasks.
Conducted comparative runtime analysis on GPUs to ensure efficient model deployment.
Benchmarked model performance using newly annotated datasets to ensure robustness and accuracy.

Sentiment Analysis Scoring (Fairness and Bias)

Conducted bias analysis and mitigation for sentiment analysis models, focusing on gender, race, and region disparities.
Utilized tools like Aequitas and IBM AI Fairness 360 (AIF360) for fairness assessment.
Implemented visualizations to display bias distribution and collaborated on model improvements to ensure equitable outcomes.

MDU Consumerism

Project Focus: Extracting insights from consumer reviews through advanced NLP techniques.
Approach: Utilized DeBERTa, fine-tuned for accuracy in aligning multiple features and opinions within long sentences of complex reviews.
Achievements: Enhanced model precision for extracting multiple relevant features and opinions.
Improved quality of insights by ensuring accurate pairing of features and opinions.
Addressed challenges associated with the intricacies of long-form text.

• Developed an AI-based consultation application to assist patients in self-diagnosing and managing acne, using TensorFlow CNN classification models to determine acne type and severity.
• Enhanced patient care by integrating AI to match patients with appropriate dermatologists based on diagnosis and track treatment progress.
• Implemented TensorFlow Object Detection for accurate lesion detection, improving the reliability of the diagnostic tool.

Clinical NER Projects
Data Scientist, Healthcare NLP
[April 2019] - [June 2020]

• Gained hands-on experience with John Snow Labs' Spark-NLP healthcare algorithms.
• Executed Clinical-NER for extracting ICD-M10 codes, SNOMED entities, and various medical code mappings, along with signs and symptoms from clinical text.
• Provided crucial insights for healthcare providers by accurately identifying and extracting medical entities.

Aiklu Triage Project
Lead Data Scientist
[June 2020] - [Dec 2022]

• Spearheading the R&D of the 'Smart Clinic by Aiklu,' aimed at digitizing dermatology clinic consultations in the United Kingdom.
• Pioneered the development of a Smart Mirror for patient registration, and a system for digital prescription processing and treatment delivery.
• Responsible for transforming traditional clinical environments into sophisticated digital platforms, enhancing patient engagement, and streamlining healthcare processes.

• Identified and resolved complex issues, providing effective solutions, and enhancing system functionalities.
• Crafted detailed project plans and communicated deadlines to ensure timely completion of milestones.
• Adhered to health and safety protocols, including social distancing and personal protective measures.
• Demonstrated flexibility by working nights, weekends, and holidays to meet project demands and critical deadlines.

• Languages: Python.
• Frameworks/Tools: TensorFlow, Spark-NLP, John Snow Labs.
• Techniques: Machine Learning, Deep Learning, Object Detection, Clinical Natural Language Processing.

• Developed specific Adult & Pediatrics Skin Condition Diagnosis Tool using TensorFlow framework.
• Worked on Tensorflow for image training and classification(CNN) Models.
• Developed team communications and information for meetings.
• Quickly learned new skills and applied them to daily tasks, improving efficiency and productivity.
• Developed and implemented performance improvement strategies and plans to promote continuous improvement.
• Delivered service to customer locations(Techion Group Limited) within specific timeframes.
• Created plans and communicated deadlines to complete projects on time.
• Collaborated with team members to achieve target results.