Working on Ultra Disk at Microsoft Azure, building ultra-low-latency, high-performance storage infrastructure at scale. Focused on latency paths, systems efficiency, and reliability for performance-critical workloads.

• Designed and implemented a GenAI generation pipeline using RAG in Azure, enabling consultants to respond to client actuary and PBM proposals faster.

• Engineered SQL data pipelines for a pharmacy recommendation system, leveraging Washington State pharmacy claims data. Used clever feature engineering techniques, to aim to aid consumers in making more informed pharmacy choices.

• I played a key role in the research and successful pilot implementation of Mettl's SpeechX suite product. This product, designed to identify Mother Tongue Influence in spoken English, generated millions of dollars in revenues.

• Our team developed innovative hybrid systems and testing methodologies that significantly streamlined the process of identifying candidates with pronounced accents. This advancement was particularly beneficial for business solution companies such as GenPact, EXL, and Hexaware. By providing a more efficient and accurate tool for evaluating potential hires' language proficiency, we enhanced their hiring processes.

Worked on developing an efficient real-time cheating detection pipeline using principles of face detection, gaze estimation, and outlier detection. This enhanced the cheating detection capabilities of the Mettl Suite and reduced cheating cases by 10%.

• Developed a BERT-based language model for the accurate identification of Hindi-English code-mixed languages in tweets. By utilizing a large-scale, multilingual dataset, we enhanced model robustness and performance by more than 2%.

• Leveraged NLP techniques, including subword tokenization (BPE and WordPiece) and RoBERTa optimization, to achieve state-of-the-art performance metrics in Hindi-English code-switched language identification.