Naga Bhuvanesh Peddi

Project 1: CPQ Platform Agentic Automation

• Designed and built an automated AI application to handle complex, multi-step form-filling tasks, significantly reducing manual data entry for enterprise Configure, Price, Quote (CPQ) platforms.
• Built out a multi-agent architecture using the LangGraph framework, mapping out the specific routing logic, state memory, and individual responsibilities for each AI agent in the workflow.
• Integrated Vertex AI and Gemini models to process and extract structured information from various unstructured customer inputs, including raw emails, PDF attachments, and CRM notes.
• Cut down manual processing time by 85% by creating a central "Rule Book" logic system that guides the AI agents on how to correctly map extracted customer requirements to the right fields.
• Set up Copilot Kit to connect the AI backend directly with the frontend user interface, allowing the application to auto-populate fields and perform actions right on the sales screen.
• Configured PostgreSQL databases to act as the stateful memory layer, ensuring the agents remembered context and user inputs across long, multi-step sales configuration sessions.
• Wrote the core backend application layer in Python using FastAPI, creating a highly responsive server to manage concurrent requests between the AI models and the CPQ frontend.
• Programmed specialized AI agents for distinct tasks, such as a dedicated "Validator Agent" that double-checks customer inputs against strict business rules before allowing the form to submit.
• Applied TDD practices to write robust unit tests for the AI's reasoning engine, ensuring accurate mapping behavior even when handling unusual edge-case customer requests.
• Created CI/CD pipelines using GitHub Actions to automate the testing and deployment processes, keeping the development and production environments completely synchronized.
• Ran extensive Manual and Automation testing alongside automated scripts to verify that the AI agents were accurately auto filling the CPQ data without generating errors or hallucinated values.

Project 2: Predictive Analysis Server Monitoring

• Worked as a Support Developer on a data-heavy predictive maintenance project designed to monitor HP ProLiant server infrastructure.
• Set up a pipeline using HP iLO (Integrated Lights-Out) to pull real-time hardware metrics and SMART sensor data from the servers.
• Used Kafka to handle high-speed streaming data, taking in server logs, system events, and text streams for real-time monitoring.
• Managed time-series telemetry data by storing it in InfluxDB, making it easier to analyze historical trends and set up threshold alerts.
• Trained Isolation Forest and LSTM (Long Short-Term Memory) machine learning models to analyze the data and predict when server components were likely to fail.
• Put together an interactive dashboard using Streamlit to give the IT team a live view of server health and predicted failure times.
• Wrote an automated notification service that instantly alerts the support team whenever hardware metrics cross safe thresholds.

Project 3: Ticket-Genie Agentic RCA Tool

• Built Ticket-Genie, an AI-powered Root Cause Analysis (RCA) tool designed to help support teams resolve technical issues faster.
• Wrote a multi-agent backend using LangGraph and FastAPI to read and analyze historical Jira tickets and system logs to find fixes.
• Set up ChromaDB as the core vector database, using hybrid search (semantic and keyword) to accurately match new issues with similar past tickets.
• Used GPT-4o alongside LangGraph to automatically read the retrieved tickets and generate clear, step-by-step resolution summaries for the team.
• Connected ngrok Webhooks directly to the AI agent so that anytime a Jira ticket was updated, the new information was instantly embedded and saved into ChromaDB.
• Brought the tool directly to the support team by connecting it to Microsoft Teams, using a Service Account to run it securely as a chatbot.
• Designed a specific "Search Agent" to query the vector database and a separate "Synthesizer Agent" to write out the final RCA reports.
• Wrote comprehensive unit tests for each agent to make sure the search results stayed accurate as the codebase grew.
• Used FastAPI to handle the communication layer between the Microsoft Teams chat interface and the LangGraph backend logic.
• Managed the initial data load of historical Jira tickets into ChromaDB and wrote scripts to keep the embeddings updated daily.
• Created a clean landing page inside the Teams app so users could easily click a button and trigger an RCA search on demand.

Project 4: Genesys Cloud CX Automation

• Currently driving the technical development of an Agentic Virtual Assistant aimed at fully automating voice and chat interactions within a high-volume Genesys Cloud CX contact center.
• Designing a Large Action Model (LAM) framework that allows the virtual assistant to actually reason through user requests and dynamically call specific tools to solve problems, rather than following a rigid script.
• Built the backend architecture ("Block 1") by developing custom REST API Action Hooks in Azure, effectively giving the AI the "hands" it needs to perform backend tasks like password resets.
• Configured the frontend intelligence ("Block 2") by creating the primary Architect Flow and the LAM "brain" directly inside the Genesys AI Studio, mapping it to our custom Azure tools.
• Established a seamless, low-latency integration bridge between the Azure-hosted backend APIs and the Genesys platform so the AI can execute tasks in real-time while on a live call.
• Programmed the system to handle natural, unpredictable human conversations, allowing the virtual assistant to understand intent and reason out the next best step without breaking the flow.
• Wrote custom REST API hooks that securely access backend databases to perform live account lookups and authenticate caller identities directly during active voice interactions.
• Working closely with business stakeholders and product owners to translate complex company policies and customer service logic into functional pathways within the Genesys Architect platform.
• Maintaining the Azure Cloud infrastructure that hosts the required APIs, ensuring the servers remain highly available and responsive enough to handle live, voice-driven tasks without lag.
• Continuously refining the AI models to successfully navigate multi-turn conversations, especially when a caller suddenly changes their mind or shifts their intent mid-sentence.