AMEY SANJAYKUMAR PANDIT

• Integrated Freshdesk with LLM-powered agents using MCPs (Model Context Protocols) to enable context-aware ticket creation, customer lookup, and automated resolution flows, reducing manual intervention in support workflows.
• Implemented Botsonic-powered conversational agents, customizing intents, knowledge bases, and tool integrations for personalized customer support experiences across web.
• Designed agentic workflows combining multiple LLMs and external tools (Freshdesk APIs, databases, internal APIs) to deliver dynamic, multi-step task automation — improving support team efficiency by 35%.
• Led the design and deployment of an automated ticketing system, leveraging NLP techniques (text classification, sentiment analysis) to reduce operational costs and accelerate resolution time by 30%, directly improving customer satisfaction.
• Built a Hospital Insurance Automation System using Streamlit, Google OCR, and GPT-based models, automating claim form processing and reducing manual effort by 40%.
• Developed intelligent client-support chatbots with advanced NLP (intent recognition, entity extraction), improving response accuracy and reducing human agent workload by 25%.
• Designed and deployed LLM-powered applications on AWS, ensuring reliability, scalability, and cost optimization with observability tools (Prometheus, logging pipelines, cron jobs for API health monitoring).
• Engineered prompt optimization and evaluation pipelines for LLMs, incorporating performance analysis scripts, A/B testing, and fallback strategies to ensure robustness and accuracy in production.
• Contributed to NL2SQL feature development, enabling natural language to SQL conversions, thereby simplifying complex data querying for non-technical users.
• Developed an Interview Q&A Generation tool using LangChain, LLMs, and vector databases for context-aware retrieval, showcasing expertise in multi-model orchestration and applied generative AI.

• Developed "iPrescribe":a personalized large scale recommender system to increase the Click-Through-Rate (CRT) for e-commerce website users utilizing deep learning models such as LSTM.
• Published papers on efficient self-learning, metareinforcement learning algorithms, and more.
• Presented research findings at national and international conferences, raising the profile of the institution.
• Managed multiple research projects simultaneously, ensuring timely completion within budget constraints.

GPA: 3.32/4 GPA

GPA: 9/10 CGPA