Somil Agrawal

Client - Union Bank of Switzerland(UBS)

• Project: Development of Testing Framework for Agentic AI in Banking Solutions [March 2025-Current]
• Technologies: Python, Langchain Framework, Google Gemini LLM model, Langraph for Agents.

Responsibilities:

• Developed a comprehensive testing framework to evaluate the performance and decision-making capabilities of an AI agent used in a banking solution, ensuring adherence to industry standards and best practices.
• Focused on key testing parameters such as Memory Coherence & Retrieval, Reasoning Relevancy & Coherence, Agent Decision Making, and Performance to ensure the AI’s operational effectiveness and compliance.
• Created evaluation metrics for measuring:
  Tool Usage: Ensured correct function/tool invocation, argument accuracy, and execution reliability.
  Task Completion & Efficacy: Verified successful goal accomplishment and response accuracy.
  Reasoning Completeness & Correctness: Assessed the agent’s reasoning process for consistency, relevance, and correctness in decision-making.
  Memory Coherence: Evaluated the agent’s ability to maintain context, retrieve data efficiently, and avoid memory degradation.
• Implemented Performance Benchmarks:
  Measured Response Time and Throughput, ensuring low-latency and high-capacity handling of multiple requests.
  Assessed Resource Utilization to optimize underlying system resources during agent operations.
• Evaluated Agent’s Decision-Making Capabilities:
  Assessed Route Selection Accuracy to ensure the agent routes inputs to the appropriate sub-modules based on user intent.
  Evaluated Subtask Coverage to guarantee the orchestrator agent addresses all identified subtasks effectively.
• Applied Agentic AI Testing Framework using a combination of custom-built tools and third-party software to simulate real-time user interactions, measure efficiency, and ensure compliance with safety and ethical guidelines.
• Led the testing and debugging efforts to improve AI’s reasoning trace coherence, ensuring the AI agent followed a consistent and logical flow in multi-step decision-making processes.
• Contributed to the enhancement of strategic planning capabilities within the AI, ensuring it could design and execute effective plans based on user input.
• Developed actionable insights from testing results, providing recommendations for refining the agent's AI-driven logic and performance optimization.
• Collaborated with clients to determine need and devise appropriate solutions.
• Delivered presentations to senior management on key project milestones, providing transparency into progress toward objectives.

Client - Union Bank of Switzerland (UBS)

• Project - AI-Driven Log Analysis and Data Interpretation. [June 2024-Jan 2025]
• Technologies: OpenAI API, Visual Studio Code, Python,Langchain, OpenAI GPT4 T LLM, Microsoft Azure OpenAI, FIX Protocol, Streamlit, RAG

Responsibilities:

• Developed a GenAI-powered chatbot using Python, Streamlit, RAG (Retrieval-Augmented Generation), and Azure OpenAI to assist with user queries in a financial services context at UBS. The solution automated key tasks in the workflow trading process, reducing manual effort and enhancing operational efficiency. Leveraged Document Intelligence to generate compliance-ready text reports and used Log Analytics to create test cases aligned with real business scenarios. Additionally, built a tool to convert raw FIX session messages into UBS-compatible formats, improving test coverage and data accuracy for system validation.
• Conducted advanced log and transaction analysis to extract, classify, and interpret financial transaction data, improving traceability and transparency in client order statuses.
• Developed structured transaction narratives and compiled a detailed workflow comparison report, achieving 60–65% accuracy across ~200 transaction records.
• Engineered synthetic test data generation pipelines and wrote BDD test scenarios (Gherkin) based on application logs, contributing to enhanced testing coverage and automation.
• Translated raw FIX protocol messages into system-compliant formats (A4, Client Sapphire), increasing message interpretability and operational readiness for OMS (Order Management Systems).
• Designed mapping documentation and data dictionaries to streamline log classification and FIX message decoding, improving cross-functional team understanding.
• Created test scenarios to validate message flows and enhance overall message coverage and validation accuracy by 60–65%.
• Collaborated with cross-functional stakeholders to validate test data integrity and improve tooling for message transformation and quality assurance.

Client - Cisco

• Project - Robotic Process Automation Project [Dec 2023- Mar 2024]
• Technologies: UiPath, Power Automate, AssistEdge, Automation Anywhere

Responsibilities:

• Trained in leading RPA tools including UiPath, Microsoft Power Automate, AssistEdge, and Automation Anywhere, with hands-on experience in designing and implementing automation workflows for business process optimization.
• Developed and tested end-to-end automation scripts for repetitive and rule-based tasks, improving operational efficiency and minimizing human error.
• Created reusable automation components and bots to streamline processes across departments such as finance, HR, and operations.
• Designed and deployed attended and unattended bots, enabling task automation across desktop and cloud environments.
• Gained experience in workflow orchestration, exception handling, and bot performance monitoring using integrated dashboards and logs.
• Collaborated with cross-functional teams to analyze business requirements and translate them into robust automation solutions.
• Improved process efficiency by designing and implementing RPA solutions for various business functions.

Infosys Validation Solutions (IVS) [Mar 2023- Dec 2023]

• Technologies: Java, Selenium, BDD (Cucumber), SOAP UI, Postman, Maven,Jenkins, SQL.

Responsibilities:

• Participated in the end-to-end software Automation testing lifecycle across web-based and API-driven applications, ensuring product quality and compliance with business requirements.
• Developed and executed automated test scripts using Selenium and Java, contributing to faster and more reliable regression and functional testing.
• Utilized BDD frameworks (Cucumber) to write clear and behavior-driven test cases, enhancing collaboration between QA and development teams.
• Conducted API testing using SOAP UI and Postman, validating RESTful and SOAP services for performance, accuracy, and security.
• Assisted in test planning, scenario creation, and defect tracking, ensuring timely issue resolution and alignment with sprint goals.
• Worked with Maven for project build management and integrated test suites into CI/CD pipelines for continuous testing.
• Collaborated in Agile/Scrum environments, contributing to daily standups, sprint planning, and retrospectives to align QA efforts with development timelines.

Infosys Validation Solutions (IVS) [Aug 2022- Feb 2023]

• Technologies: Java, Selenium, BDD (Cucumber), SOAP UI, Postman, Maven,Jenkins, SQL.

Responsibilities:

• Completed an intensive training program focused on Automation Testing, API Testing, and Database Management, gaining hands-on experience in Java, Selenium, SOAPUI, SQL, UNIX, and Hibernate.
• Tested web applications and REST APIs, ensuring functionality, performance, and security through automated scripts and manual validation.
• Developed a capstone project that involved testing web applications and APIs, applying industry best practices and automation techniques to streamline test processes.
• Gained proficiency in using SQL for database querying and validation, and UNIX for managing environments and executing test cases.
• Enhanced business communication skills through training, enabling effective collaboration with cross-functional teams.
• Successfully cleared the training exam with a score of 70%, demonstrating a solid understanding of automation and testing methodologies.