Headline

RL Lead @ Deccan AI | Post training, SFT, RLHF, LLM Evals, RL Gym

About

AI Architect | Full Stack AI | LLM | MLOps | AI Agents | RAG With a solid foundation in data science and over 9+ years of experience, I excel in devising and deploying innovative AI solutions across a spectrum of industries such as e-commerce, agri-tech, and healthcare. My expertise encompasses a broad range of technologies including computer vision, natural language processing, machine learning, deep learning, Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), and Generative AI. I adeptly apply these technologies to complex problems, achieving breakthroughs in classification, detection, segmentation, and generative tasks.

Location

Hyderabad

Work experience

Agentic Lead at Deccan AI, Mar 2026 - Present · 7 mos, Hybrid, After building internal pipelines, now leading everything related to coding, RL , agentic client fulfilments: - Leading E2E projects from data collection process, platform development to training models for client data shipment.

ML Lead at Deccan AI, Aug 2025 - Apr 2026 · 9 mos, - Built multiple features to streamline the sourcing, assessment and fitment.

Data science at Mensa Brands, Nov 2022 - Aug 2024 · 1 yr 10 mos, Abu Dhabi, Abu Dhabi Emirate, United Arab Emirates, Data science | MLOps

Principal Data Scientist at Johnson Controls, Feb 2021 - Aug 2022 · 1 yr 7 mos, Bangalore Urban, Karnataka, India, - Engineered a multi-tenant architecture for the business logic layer, dramatically reducing inference time by 10x, and significantly enhancing system efficiency and performance. - Advanced the object detection system from single-class to multi-class detection and tracking, broadening the module's capabilities and application scope. - Acted as a pivotal member of the design committee for a next-generation AI platform, contributing to strategic planning and innovative solutions.

Data Scientist at MFine, Mar 2019 - Mar 2021 · 2 yrs 1 mo, Bengaluru, Karnataka, India, - Designed and developed a MLOps infrastructure from scratch, optimizing AI model management, resulting in a 50% reduction in re-training time and a 75% decrease in AWS costs. Achieved rapid production deployment and comprehensive tracking, shortening the POC to production cycle to just 3 days, utilizing Metaflow and AWS Services for complete ML lifecycle management. - Developed AI models to automate case report section suggestions for junior doctors, reducing their workload by 20% and enhancing report accuracy. - Contributed to a variety of projects, including fundus grading, development of an ML visualization library, X-Ray analytics, and initiatives to optimize doctor workload, demonstrating versatility and impact across healthcare AI applications.

Data Scientist - IV at SigTuple, Dec 2018 - Mar 2019 · 4 mos, Bengaluru Area, India, - Led the OCT analysis clinical study for DME classification, achieving key milestones in medical imaging research. - Developed and managed the Dipstick Analyzer project, overseeing everything from synthetic data collection to model development for 7-parameter analysis, including innovative image normalization and clustering algorithms.

Senior Computer Scientist at SigTuple, Feb 2018 - Dec 2018 · 11 mos, Bengaluru, Karnataka, India, - Revolutionized DR grading with a novel CNN, enhancing classification and detection, leading to two successful clinical studies and AI service deployment. - Led DME OCT scan classification, achieving a clinical study milestone before strategic discontinuation. - Obtained a patent for a groundbreaking enhancement in CNN architecture for DR/OCT analysis. - Integrated Unet architecture into the AI platform, enabling advanced segmentation model training.

Computer Scientist at SigTuple, Mar 2017 - Jan 2018 · 11 mos, Bengaluru, Karnataka, India, - Revamped WBC extraction with a Computer Vision algorithm, boosting recall from 60% to 98%, pivotal in a major clinical trial's success. Published findings at MICCAI 2017. - Overhauled WBC screening and classification models, employing diverse architectures and visualization techniques to rectify misclassifications and enhance data quality, resulting in a 34% performance increase. - Led rigorous testing and optimization of 100+ model variations, achieving a significant leap in accuracy to 96%, demonstrating a committed, iterative approach to model enhancement.

Education

Bachelor of Technology (B.Tech.), Electrical and Electronics Engineering, National Institute of Technology Karnataka

Karnataka Pre University Board, Applied Sciences, Diamond PU Science College