Rahul Kamath Email & Phone Number

Bengaluru, Ka, In

Chennai, Tn, In

Bengaluru, Karnataka, India

• Fine-tuned Large Language Models (LLM’s) on clinical text implementing advanced scaling techniques, including model parallelism, data parallelism and pipeline parallelism to automate complex medical coding tasks.• Developed and deployed ML models for insurance claims denial prevention and management which is projected to decrease BPO tasks and spends by 20-25% resulting in savings of approx. $5Mn/year. • Investigated and addressed issues with production ML systems by systematically analyzing metrics and logs across different production system components including our prediction service and prediction store.

Bengaluru, Karnataka, India

• Member of athena intelligence (athenahealth’s R&D division)• Medical coding using Electronic Health Records (EHR)• Insurance claim denial prediction and management for Revenue Cycle Management team

Bangalore Urban, Karnataka, India

Led a team of 4 data scientists in managing 20+ ML models in production for the analytics team of Microsoft Azure including model enhancements, prod. issue resolution, migration to cloud and performance monitoring.

Bangalore Urban, Karnataka, India

• Leveraged Bayesian models, tree-based models and survival analysis to identify drivers of sales erosion for a large Swiss pharmaceutical company leading to 30% increased chance of non-erosion and 3-4 months delay in erosion.• Modelled the propensity of a physician adopting a new drug using Gradient Boosting technique providing 4x lift over a random model and 1.5x lift over traditional rule-based approach.• Managed vendor implementation of a web application showcasing 15+ ML use cases which was very well received by partners across the company and clients across the globe.

• Built a next generation AI-powered people search engine targeted at the UK market which provides useful information aggregated from a variety of sources using advanced Artificial Intelligence algorithms.• Indexed more than a million professionals with detailed information obtained using distributed web crawlers and processed using Machine Learning algorithms. The service had around 20000 page-views per month at its peak.• Optimized the websites' performance which resulted in Google Pagespeed Insights score of 97/100 across devices.• The web application is built using cutting edge technologies including Containerization (Docker), Web Components (Polymer), NoSQL (MongoDB), Full-Text Search (Elasticsearch) and deployed using AWS tools like EC2, S3, CloudFront and Route53.

Tokyo, Japan

• Worked from scratch on the design and development of a people search engine SPYSEE 2 (http://spysee2.jp), which automatically collects and visualizes information with a strong focus on human relations.• Developed contemporary Machine Learning and Natural Language Processing based algorithms for the extraction of various types of information about people (bio-data, profile picture, social network, related webpages etc.) from unstructured and noisy web data.• Developed a solution for name disambiguation problems in short texts using a probabilistic classifier.• Implemented a high-performing face recognition solution using a combination of Google’s OpenFace library and K-Means clustering.• Crawled millions of webpages efficiently using programs built to run using Amazon Web Services.• Solely responsible for maintaining the legacy site SPYSEE (http://spysee.jp), which at one point was Japan's most popular people search engine which had more than 20 million page-views per month at its peak.

Tokyo, Japan

• Developed modules capable of extracting a person's social network from the web using heuristic techniques.• Developed a face detection module using OpenCV capable of identifying faces of people from images with a high degree of accuracy and at scale.• Developed a module to identify people trending on news websites using Klienberg's burst detection algorithm.

Master Of Engineering - Meng, Ml & Nlp

Bachelor Of Engineering - Beng