Bengaluru, Karnataka, India
I'm a passionate Data Scientist with a strong background in mathematics, machine learning, and AI, currently with the AI Forward Deployed Engineering (AI FDE) team at Databricks. I’m driven by tackling complex problems and creating impactful, scalable solutions across industries. Whether it's optimizing demand forecasting at Walmart, designing anomaly detection systems at AIS, or now deploying cutting-edge GenAI applications at Databricks, I’m always focused on delivering tangible business value. What I do - Machine Learning & Data Science: I have a proven track record in building end-to-end data pipelines, deploying production-grade ML models, and implementing MLOps practices across a wide variety of industries. - AI Solutions & GenAI: Currently focused on developing and productionizing GenAI applications (RAG, fine-tuning, multi-agent systems) and working closely with customers to drive AI innovation. - Collaborative Leadership: I thrive on working cross-functionally with product, engineering, and customer teams to shape strategic AI priorities, refine technical solutions, and push the boundaries of what’s possible with machine learning and AI technologies. Key Highlights - Optimized and deployed ML models for large-scale demand forecasting at Walmart, impacting millions of customers. - Built robust data pipelines and designed infrastructure that supports data-driven decision-making and scalable model deployment. - Contributed to MLOps processes, ensuring smooth deployment and management of models in production environments. - Designed and implemented an anomaly detection system using auto-encoder neural networks for real-time data monitoring. Core Skills - Python, PySpark, SQL, Scikit-learn, TensorFlow, PyTorch - Machine Learning, Data Science, Neural Networks, GenAI - MLOps, Kubernetes, Airflow, AWS, Azure, GCP - LLMOps, LangChain, HuggingFace, DSPy - Data Visualization, Cloud Infrastructure, Distributed Computing
Spearheaded the development of Generative AI applications, delivering innovative solutions such as: – Automated Prompt Engineering Agent: Conceptualized and implemented an agent that harnesses LLMs to dynamically optimize prompts for diverse use cases. Enhanced Developer Productivity by 98% and increased accuracy of responses by 26%. – LLM-As-A-Jury Agent: Increased accuracy for Hallucination Detection by 62% and reduced false positives by 88% through a novel jury-based framework. – LLM Prompt Compression: Analyzed and utilized LLM Lingua to compress prompts, maintaining accuracy while reducing costs. – LLM Caching, Guardrailing, and Intent Identification: Engineered and streamlined innovative solutions to boost efficiency in Generative AI workflows, including caching mechanisms, intent recognition, and safeguard integrations. – LLM-Based Anomaly Detection: Pioneered a novel anomaly detection application utilizing LLMs, eliminating reliance on use-case-specific retraining or fine-tuning of traditional models. – Post-Deployment Metrics: Architected comprehensive frameworks to evaluate and monitor LLM performance, focusing on detecting hallucinations and identifying drifts in input prompts for sustained reliability.
- Leveraging my data science expertise, optimized and deployed large-scale ML models for demand forecasting across 135,000+ Walmart Omni channel items. - Developed comprehensive infrastructure, including dashboards, alerts, data pipelines, and feature engineering, to support robust data-driven decision making. - Contributed to model registry design for efficient model lifecycle management, fostering a data-centric approach. - Engaged in low-level ML component design and collaborative architecture improvement, addressing corner cases and code gaps, demonstrating strong problem-solving skills. - Optimized neural network training times by identifying and resolving bottlenecks, showcasing a commitment to continuous improvement and efficiency.
- Developed and implementend an Anomaly Detection Service to proactively identify potential failures in SQL server usage. This involved designing and building an auto-encoder neural network, leveraging my strong mathematical background. - Worked collaboratively on the technical design and development of the end-to-end pipeline, encompassing data ingestion, model scoring, and real-time alerting for SQL server usage anomaly detection. - Leveraged Azure cloud technologies to collaboratively deploy the service with cross-functional teams, ensuring seamless integration and scalability within the clients existing cloud infrastructure.
Presented a new proof to the Isoperimetric Inequality, based on Peter D. Lax’s work on the calculus of variations.