Harish A.

- Built autonomous AI agents using LangGraph, CrewAI, and AutoGen that handle investment workflow tasks with minimal human oversight. These agents cut down manual work by 60% and made our operations much more efficient. - Created a multi-agent system that coordinates different AI tools to complete complex financial analysis tasks. The system can call functions, query APIs, and use vector databases (Pinecone and Weaviate) to maintain context across conversations, reducing analysis time by 45%. - Developed retrieval-augmented generation (RAG) systems for compliance and research teams that combine different search methods to find relevant information quickly. Spent a lot of time tuning the retrieval and reranking to get answers that made sense for banking use cases, achieving 85% accuracy on compliance queries. - Set up the infrastructure to deploy and monitor these AI systems using Docker and Kubernetes on GCP. Implemented proper Machine Learning Operations (MLOps) practices with MLflow and Kubeflow so we could test different agent versions and push updates without breaking production, reducing deployment time by 70%. - Worked on autonomous agents that handle portfolio rebalancing, risk checks, and regulatory reports by connecting to Vertex AI and BigQuery. These systems process thousands of transactions daily and include real-time fraud detection with 99.2% accuracy. - Collaborated with product managers, compliance teams, and other engineers on agent design and implementation. Established testing standards using pytest and simulation environments, which helped catch issues before they reached production, reducing production bugs by 80%. - Built safety features for our AI agents since we're in a regulated industry—things like validating outputs, detecting hallucinations with 92% precision, logging everything for audits, and ensuring we meet SOC2 and GDPR requirements.

- Built AI agent systems using LangChain and LangGraph on Azure that automate key retail operations like inventory management, pricing updates, and targeted promotions. These agents handle millions of customer interactions each month and helped boost our conversion rates by 23%. - Developed AI agents that can use tools and pull information from our knowledge bases to resolve customer service issues independently. Used Azure Cognitive Services and AutoGen to build agents that think through problems step-by-step, which cut down escalations to human agents by 40% and improved how accurately we solve customer problems to 88%. - Created ML pipelines that learn and improve over time using reinforcement learning with Ray and MLflow. The agents analyze customer feedback and adjust their recommendation strategies automatically, which led to 18% better cross-sell results and $2.3M in measurable revenue growth. - Set up the infrastructure to run these agents at scale using Docker and Kubernetes on Azure. Implemented event-driven architecture so agents can handle tasks in parallel and retrain models when needed. This dramatically shortened our deployment timelines by 65% and reduced compute costs by 35%. - Worked with product teams, engineers, and leadership to figure out how to responsibly deploy autonomous AI systems. We established testing frameworks to catch issues like hallucinations, built transparency into agent decision-making, and made sure our AI strategy aligned with business goals across multiple teams. - Implemented agentic workflows for personalized marketing campaigns that dynamically segment customers and generate tailored content using GPT-4 and Azure OpenAI. These campaigns achieved 31% higher engagement rates and reduced campaign creation time from weeks to hours.

• Designed and implemented NLP models for chatbots and virtual assistants to automate customer support, improving response accuracy by 30% for banking queries. • Deployed RAG-based solutions in production to enable real-time knowledge retrieval for chatbots, virtual assistants, and enterprise search. • Developed Transformer-based architectures (BERT, GPT variants) fine-tuned on financial datasets to extract insights from unstructured banking documents, including loan applications, credit reports, and transaction histories. • Integrated ML pipelines using TensorFlow and PyTorch to detect fraudulent transactions, analyzing large-scale, real-time banking transaction streams with anomaly detection algorithms. • Applied reinforcement learning techniques to optimize algorithmic trading strategies, improving portfolio risk management and maximizing returns. • Built risk assessment models using time series analysis and predictive modeling for credit scoring and default prediction. • Participated in cross-functional agile teams to ensure AI solutions met banking regulations and security standards. • Developed Agentic AI and LangChain workflows enabling autonomous multi-step operations such as automated loan processing and compliance checks, reducing manual intervention and improving turnaround time by 40%. • Designed and deployed advanced RAG pipelines combining pre-trained language models with dynamic retrieval from large-scale banking knowledge bases.

• Designed and implemented scalable financial data pipelines using PySpark on Dataproc and Databricks, ensuring compliance with banking regulations and data security standards. • Processed nested JSON financial data with BigQuery SQL and Spark SQL, enabling efficient querying and integration into risk and compliance analytics platforms. • Automated secure ETL workflows for ingesting and loading sensitive banking data into Cloud Storage and BigQuery for compliant data warehousing and regulatory reporting. • Built reusable libraries integrating multiple banking databases (DB2, Oracle, SQL Server, MongoDB) with Spark for diverse financial data ingestion and consolidation. • Created and maintained Cloud Dataflow / Dataproc ETL workflows to automate transformation of trading, credit, and compliance datasets with minimal manual intervention. • Architected a financial data lake on Cloud Storage to support advanced analytics and ML use cases for fraud detection, credit risk modeling, and portfolio optimization. • Collaborated with business stakeholders to gather requirements, perform data validation, lead engineering teams, and deliver production-ready data solutions aligned with banking objectives. • Acted as liaison between technical teams and banking clients to gather domain-specific requirements, conduct code reviews, and manage version control with GitHub.

Completed a rigorous graduate program with a focus on full-stack development, cloud computing, and machine learning. Specialized in building scalable enterprise applications and integrating cutting-edge technologies such as Generative AI (LLMs, GPT, LangChain) and semantic search into real-world solutions. Gained hands-on experience with AWS, DevOps pipelines, and MLOps practices through coursework and student-led projects. Led projects involving natural language processing, cloud-native microservices, and AI-powered search, applying research concepts in areas like LLM fine-tuning, prompt engineering, and vector embeddings. Actively participated in technical communities and hackathons, winning recognition for innovative use of AI in workflow automation.