Upashana Chatterjee

Full Stack Developer with 1+ year of hands-on experience at PwC India (via Infocus Technology), specializing in building agent-based systems using multi-agent frameworks. Experienced in designing autonomous, context-aware AI solutions leveraging advanced LLM capabilities, and in architecting intelligent systems that integrate seamlessly with enterprise environments while ensuring performance, reliability, and scalability. Proficient in Generative AI, Large Language Models (LLMs), and advanced prompt engineering, with deep experience building Retrieval-Augmented Generation (RAG) architectures for domain-specific, high-accuracy solutions.

Strong backend development skills in Python, PostgreSQL, FastAPI, and REST APIs, complemented by experience with modern AI frameworks such as LangChain, LangGraph, and Hugging Face, and with CI/CD pipelines supporting efficient deployment and maintenance of AI systems.

Full Stack Developer - PwC India (via Infocus Technology) - Kolkata, WB

• Built and maintained scalable enterprise web applications using React and ASP.NET for internal users across departments, developing reusable components that improved UI consistency and reduced defects across modules.
• Designed RESTful APIs backed by PostgreSQL and Spring Boot, optimising critical queries for significantly faster data retrieval and integrating backend services with React frontends to deliver complete end-to-end features.
• Developed backend services in Python, designing and consuming RESTful APIs to support an autonomous Site Reliability Engineering (SRE) system for a microservices platform, enabling continuous failure detection, AI-driven incident analysis, and automated remediation workflows with human approval gates.
• Containerized business-telecom services (authentication, billing, support, subscriptions) and SRE multi-agent workflows (monitoring, analysis, healing, orchestration) using Docker and Docker Compose, deploying and managing them on Azure Container Apps with container-orchestration best practices.
• Engineered detection and monitoring mechanisms using health checks, Log Analytics, database-listener agents interfacing with MongoDB/Redis-like systems, and chaos-simulation endpoints to ensure comprehensive incident capture.
• Integrated and orchestrated AI/ML workflows using LangChain and LangGraph, enabling modular, extensible pipeline management and advanced prompt orchestration for AI-driven incident diagnosis.
• Built an AI-powered incident-diagnosis pipeline using Azure AI Foundry as the primary LLM source and Cohere as a fallback, delivering validated root-cause analyses and remediation recommendations through a responsive React diagnostics interface.
• Developed secure authentication and authorization using JWT/session-based best practices to protect incident-approval workflows, implementing deterministic healing plans with strict approval gates and configurable human-in-the-loop modes, including one-click React-driven approvals.
• Integrated full-lifecycle incident management with ServiceNow via RESTful APIs, automating ticket creation, comments, state transitions, and closure, and maintaining audit trails and incident timelines through React dashboards.
• Conducted experiments with an XGBoost-based forecasting pipeline on six years of synthetic data to benchmark production readiness and model performance.

MCA - Meghnad Saha Institute of Technology (MAKAUT) (2022-01 - 2024-12)

BCA - Techno Main Saltlake (MAKAUT) (2019-01 - 2022-12)