As an ML (Machine Learning) hardware accelerator engineer I was mainly responsible for designing, developing, and optimizing the Hadware Abstraction Layer code such as to harness the capabilities of the Neural Processing Unit(eNPU) chip for accelerating machine learning tasks in the best possible way. Here are some of my key responsibilities as an ML hardware accelerator engineer:AI Model Development and Optimization: Led the development and optimization of AI models, focusing on… Show more As an ML (Machine Learning) hardware accelerator engineer I was mainly responsible for designing, developing, and optimizing the Hadware Abstraction Layer code such as to harness the capabilities of the Neural Processing Unit(eNPU) chip for accelerating machine learning tasks in the best possible way. Here are some of my key responsibilities as an ML hardware accelerator engineer:AI Model Development and Optimization: Led the development and optimization of AI models, focusing on enhancing performance and efficiency. Applied advanced techniques to improve model robustness and achieve high accuracy in various applications.Multimodal AI Systems: Directed efforts in designing and implementing multimodal AI systems, integrating diverse data sources such as text, images, and other modalities to enhance model accuracy and effectiveness.Collaboration and Technology Transfer: Worked collaboratively with cross-functional teams to translate advanced AI research into practical solutions, ensuring smooth integration and impactful results across product lines.Technical Proficiency and Innovation: Showcased expertise in Python and machine learning frameworks like PyTorch, with a strong foundation in deep learning and uncertainty estimation methods, driving innovation and excellence in AI projects. Show less

I was working on Self-Supervised representation learning techniques for Automatic Speech representation. I implemented wav2vec paper from scratch and did several modifications to it based on the phonemic transcriptions available at hand leading to its performance enhancement. I also explored some of the state-of-the-art ASR techniques which derived from wav2vec like wav2vec2 and wav2vecU.

Data Analysis: Analyzing sales data, customer demographics, market trends, and other relevant data to identify insights and patterns that can inform call planning and territory alignment strategies.Solution Design: Designing data-driven solutions and strategies for optimizing call planning and territory alignment, taking into account factors such as physician preferences, geographical considerations, sales goals, and regulatory requirements.Data Visualization: Creating… Show more Data Analysis: Analyzing sales data, customer demographics, market trends, and other relevant data to identify insights and patterns that can inform call planning and territory alignment strategies.Solution Design: Designing data-driven solutions and strategies for optimizing call planning and territory alignment, taking into account factors such as physician preferences, geographical considerations, sales goals, and regulatory requirements.Data Visualization: Creating visualizations, dashboards, and reports to communicate insights, performance metrics, and key findings to stakeholders and decision-makers, facilitating data-driven decision-making and strategy refinement.Performance Monitoring: Monitoring and evaluating the performance of call planning and territory alignment strategies, tracking sales metrics, customer engagement, and sales rep productivity, and identifying opportunities for improvement and optimization. Show less

Worked for development of an android app named ‘SecurePass’ which uses a real time fingerprint based identification system and is therefore highly secure reducing the possibilities of account hacking or counterfeit crimes.It was designed to be used for attendance systems and accessing accounts.

Investigated various fuzzy logic-based clustering methods, including Fuzzy C-Means (FCM), Possibilistic C-Means (PCM), Fuzzy Possibilistic C-Means (FPCM), and Possibilistic Fuzzy C-Means (PFCM).Developed and applied an interval type-2 Gaussian Kernel-Possibilistic Fuzzy C-Means clustering approach to diverse datasets and RGB images, evaluating its performance on datasets with ambiguous class boundaries.