I built DeepMindScape as an applied-AI lab focused on designing production-grade voice automation systems across multiple industries. I developed end-to-end voice pipelines — real-time ASR, LLM agents, TTS, evaluation harnesses, and monitoring — for two major verticals: healthcare and real estate.

On the healthcare side, I engineered voice agents that handled patient intake, scheduling, insurance checks, billing/coding support, and RCM workflows.

Separately, I built a second suite of voice agents for property management and leasing (check out www.JusDials.com), including multilingual leasing assistants, maintenance triage, tour scheduling, and resident support.

Both verticals operate on independent architectures tailored to their domain, but I unified them through MCP servers and tool-calling layers that integrate with PMS systems, software suites, and external APIs for actions like scheduling, ticket creation, data retrieval, and policy lookup.

This involved:

• Designing real-time voice pipelines for domain-specific intents

• Developing MCP servers to integrate with PMS/RCM/CRM systems

• Building knowledge-base ingestion + structured retrieval for each vertical

• Architecting agent behaviors, business-rules execution, and persona design

• Creating evaluation + calibration systems for call quality and reliability

• Architected a scalable software architecture of an E-commerce platform on AWS

• Developed a complex pricing algorithm to reduce the realtime browsing latency, trained on over 10M tuples of categorically sorted historical data

• Built a search indexing based chatbot to resolve platform related queries of cusotmers

• Worked on the social marketing side -- video editing, promotional content creation for the social media platforms

• Participated in connecting with vendors, onboarding them and getting the SLAs in place

• Developed an IT skills training chatbot starting with prompt engineering and OPENAI's API.

• Achieved a 94.8% improvement rate in user skill levels, measured through SentenceTransformer's training data.

• Implemented advanced Large Language Models (LLMs) such as LLaMA2 and BART with vectorization techniques to minimize API reliance.

• Constructed a chatbot utilizing LLaMa2's pre-trained model with 7 billion parameters.

• Conducted hyperparameter tuning, including temperature=0.7, top-p=0.8, top-k=40, and max_tokens=512, for optimal chatbot performance.

• Developed and implemented shell scripts to automate program execution, facilitate file manipulation, and generate text outputs, resulting in increased operational efficiency.

• Acquired proficiency in data visualization tools such as Tableau and Power BI, utilizing them to analyze and present complex data sets.

• Spearheaded the creation of a comprehensive sales dashboard on Tableau, enabling real-time monitoring and analysis of product sales performance, contributing to informed decision-making processes within the company.

• Organized multiple faculty development programs(FDPs) on fabrication

• Supervised several projects and guided students to build a robust system for their projects and startups