Vinay Araveeti

Full Stack LLM Development Associate Manager with expertise in GenAI, Agentic AI, RAG, and intelligent automation - building AI-powered applications across Quality Engineering, Payments, Data Management, and Enterprise Productivity domains. Transformed complex business challenges into innovative products using Python, Flask, FastAPI, embeddings, and LLM technologies, improving efficiency and aiding modernization.Few of my works as below

• SkillForge - Transforms enterprise documents into versioned skill/context files via NLP + LLM pipeline with human-in-the-loop keyword curation.Reduced LLM token consumption by ~30% per query (1,500 → 1,000 tokens) by replacing raw-document RAG with distilled skill-file RAG - same accuracy, 0 points lost - saving 338.7M tokens per year at a 50-agent fleet scale.
  Technologies: Python ·Flask ·SQLite ·LLM ·spaCy ·YAKE ·TF-IDF ·Clustering techniques(K-Means)
• ClauseIQ - Contract/document intelligence with hybrid search and RAG-based answer generation.
  Technologies: Python ·FastAPI ·SQLite ·Embeddings ·Hybrid search ·LLM
• PayGenLab - SWIFT payment journey simulator with deterministic multi-hop chain transformation and message template generation.Eliminated manual SWIFT message construction across 32 corridors and 7 message types, generating syntactically correct synthetic payment messages-STP, AML, duplicate, currency mismatch and more in seconds instead of ~30 minutes per manually crafted test fixture.
  Technologies: Python ·Flask ·SQLite ·LLM ·SWIFT schemas (MT103/MT202/pacs.008)

Architect role:

• Responsible for designing and enhancing automation frameworks, integrating multi-tool ecosystems (UFT, Selenium, Postman, APIs), and developing SDLC productivity accelerators that enable scalable end-to-end automation and delivery efficiency.

Research and Dev activities:

• Quick Data Assist: Developed a web-based modernization solution that simplified complex IMS Screens navigation for data discovery. The solution evolved from a database-driven search utility into a reusable Data API platform, and subsequently into an AI-powered Test Data Agent for intelligent test-data discovery, retrieval, and provisioning across QA workflows.Reduced data discovery effort by ~1 hour per user per day through intelligent search. Web UI → Data API → Test Data Agent

Technologies: Java ·JSP ·HTML ·CSS ·jQuery ·JDBC ·IMS driver

• Evidence Report Consolidator: Python desktop tool built for a Sustainability initiative that analyzes failed and redundant automation framework output reports, generates a consolidation summary, and lets users selectively delete flagged files to reclaim storage.Helped in reclaiming up to 50GB of storage per release cycle.

Technologies: Python ·Tkinter ·ttkbootstrap

• Mainframe Screen Modernization PoC: Led a proof-of-concept to modernize GBS mainframe screens into a web-based UI using IBM IDz, exposing APIs to improve developer and SDLC user experience across mainframe workflows.This helped in reducing execution time from 10 minutes to under 1 minute (>90% improvement).

Technologies: IBM IDz ·Mainframe ·REST APIs ·Web UI

Leadership role:

• Coordinating with clients on open positions and staffing process.
• Developed 'Buddy Program' learning framework to facilitate transition of 50+ functional testers to techno-functional roles, enhancing their skills in UFT and Tosca automation tools.
• Coordinated review meetings on automation metrics, providing guidance for feasibility analysis and advancing In Sprint Automation initiatives.

Architect role:

• Analyzed requirements and modified automation framework, integrating tools like Postman, UFT, and Selenium to achieve end-to-end flow automation.
• Groomed new recruits and freshers on automation tools and framework to facilitate seamless transformation.
• Conducted regular meetings with team members to address issues and provide necessary guidance.

Functional Testing role:

• SIT and Sprint Testing application for outgoing messages includes 700 series messages (765,767/775,768,769,760/761). Accounting validations for various payments.
• Non payments like FX(Foreign Exchange) and TD(Time Deposit) screen and DB validations.

Research and Dev activities:

• Cucumber + LeanFT framework PoC: Developed a LeanFT, Cucumber, and TestNG automation framework that reduced execution time by 80% and improved framework maintainability.This framework is more lightweight allowing for quicker execution and reduces resource consumption compared to UFT's heavier model.
• Utility to trigger UFT on multiple citrix machines on single go: Built a web utility using socket programming to trigger UFT execution across multiple Citrix machines with a single click, enabling parallel execution and reducing manual effort
• Implementation of clever Login and Logout process: Developed a centralized login/logout framework using SQL-managed automation IDs, secure authentication, and session management to improve automation reliability and execution efficiency.

Functional Testing role:

• Scope of the project is to test the incoming and outgoing telegraphic transfers (OTT, ITT), inward and outward ACH & RTGS(INT,ILT,ONT,OLT) payments include the payment processing, charge validation, MT(MT103,202,202COVER) and MX(Pacs.003, Pacs.004, Pacs.008,Pacs.009) message validations, report validations. Conducted SIT testing and developed regression packs.
• Framework for Shift left Automation: Designed a hybrid shift-left automation framework combining record-and-replay and customized UFT components, enabling collaboration between functional and automation teams and earning a Client Best Performer Award.
• Swift Message Automation: Built One Touch Automation, reducing SWIFT message setup and processing time by ~30 minutes per test case through one-click automation.
• Framework for Shift left Automation: Built an award-winning hybrid automation framework using UFT RegisterUserFunc, bridging functional and automation testing through reusable custom components.

Automation Framework developer role:

• Developed in-house hybrid framework for UFT, integrating data-driven, modular, keyword-driven, and non-OR model frameworks as part of CoE team.

Research and Dev activities:

• UI for Automating Automation: Created .NET-based UI to generate automation framework scripts with a single click, leveraging OTA of UFT and ALM.
• Hybrid OR model framework: Engineered hybrid OR model framework to combine benefits of both OR and non-OR methodologies.
• Pre-Requisite Data Retrieval: Using ODBC connectivity, using VB Script retrieved required data from database reducing manual input efforts.