Nitin Panwar

I'm working as a full-stack engineer with a strong back-end focus in Python, building software systems to manage and monitor distributed energy container infrastructure. My role involves designing and developing scalable microservices in Python and Node.js that handle end-to-end real-time telemetry, health monitoring, and performance analytics, along with React-based dashboards for operational visibility.

I'm currently leading a proof-of-concept implementing agentic AI pipelines using OpenAI and Gemini models to support rack-architecture troubleshooting. The solution leverages engineering knowledge to assist field technicians through contextual diagnostics, automated issue transcription, and intelligent resolution suggestions, improving operational efficiency and reducing manual intervention.

This project involved the development of a scalable video streaming platform, designed to facilitate live content broadcasting. The platform enabled content creators to conduct one-way audio/video live sessions with audiences, augmented by real-time viewer engagement features such as reactions and live chat functionalities. The initial implementation utilized AWS Chime, which was subsequently migrated to the Agora SDK to enhance feature capabilities and streaming performance.

A key enhancement to the platform included the integration of a cloud recording feature, ensuring that live sessions are stored as HLS streams. This content is then made available for on-demand replay via a robust CDN network, significantly expanding content accessibility and extending its lifecycle beyond the live broadcast.

I designed and built an AI-powered IVR calling system for a HIPAA-compliant healthcare platform, focused on delivering natural, emotionally intelligent patient interactions at scale. The system leveraged OpenAI APIs for agentic, multi-turn conversations and Whisper for real-time speech-to-text conversion, enabling accurate understanding of patient responses. Text-to-speech pipelines were tuned to produce human-like, empathetic voice responses, while validation and safety layers ensured clinical relevance, emotional appropriateness, and regulatory compliance throughout each interaction.

The IVR architecture was implemented using Python-based microservices, with Flask orchestrating AI inference, conversation state, and call scheduling workflows. Django Admin provided secure governance, auditing, and controlled access to sensitive conversational data. Conversational signals and emotional cues were analyzed to refine AI behavior and improve engagement quality continuously. This solution enabled scalable, intelligent patient outreach while aligning with research objectives and ethical healthcare standards.

I built an AI-powered automated property listing engine for a property discovery platform, designed to generate accurate, high-quality listings from uploaded property images. The system leveraged existing property datasets and historical listings through an RAG pipeline, grounding AI-generated metadata with verified attributes such as location, layout, amenities, and pricing patterns. Advanced image-matching techniques were used to compare uploaded images with existing listings, enabling feature extraction, similarity scoring, and duplicate detection. Sanity checks and validation layers ensured the generated listings remained consistent, realistic, and aligned with market constraints before being published.

To improve property discovery, I implemented agent-based workflows that translated natural language user intent into structured database filters and search queries. These agents enabled users to describe requirements conversationally, which the system converted into precise search criteria across the property catalog. The AI-assisted listing and search experience significantly reduced manual effort, improved listing accuracy, and made property discovery more intuitive for both agents and end users.