---

👨🏻‍💻 About Me

📄 More about me

• 🎓 M.S. in Information Systems (Data Science) — University of Maryland, Baltimore County
• 🎓 B.E. in Computer Engineering — Savitribai Phule Pune University
• ⚙️ Comfortable across Agile and Waterfall; happiest taking something from a blank repo to production.

---

⚡ Now — October 2026

• 💡 Shipped Lampline v0.1.0, my self-hosted voice agent platform. Next up: one-command Google Cloud deployment and Genesys Cloud AudioHook, so contact centers can hand calls to self-hosted agents.
• 🚗 ioniq6-pilot: first real-car tests of sentry mode and steering-tune feedback.
• 🤝 Open to conversations about voice AI, contact-center automation and self-hosted LLM agents.

---

⚙️ Tech Stack

Languages

Conversational AI & VUI

GenAI & NLP

Web & App

Cloud, Data & Tools

---

🏗️ How a call flows

The voice-agent architecture I design — in production on Dialogflow CX + Genesys, and self-hosted with Lampline.

flowchart LR
  caller(["📞 Caller"])

  subgraph tel["Telephony"]
    direction TB
    genesys["Genesys CX"]
    sip["Twilio · Telnyx · SIP<br/>via jambonz"]
  end

  subgraph runtime["Agent runtime — Dialogflow CX or Lampline"]
    direction TB
    stt["🎙️ Speech-to-text"]
    router{"Flow / page<br/>router"}
    llm["🧠 LLM step<br/>Claude · Gemini · GPT"]
    scripted["📋 Scripted page<br/>forms · rules"]
    tts["🔊 Text-to-speech"]
    stt --> router
    router --> llm
    router --> scripted
    llm --> tts
    scripted --> tts
  end

  tools[("🔌 Webhooks & MCP tools<br/>billing · records · CRM")]
  human["👩‍💼 Live agent"]

  caller <--> tel
  tel --> stt
  tts --> tel
  llm <--> tools
  scripted <--> tools
  router -. "handoff" .-> human

  classDef accent fill:#F75C7E,stroke:#F75C7E,color:#ffffff
  classDef brain fill:#7C3AED,stroke:#7C3AED,color:#ffffff
  class router accent
  class llm,scripted brain

Loading

---

🚀 Featured Projects

A mix of what I'm building in the open and the conversational-AI systems I've shipped in production.

💡 Lampline — flagship, open source

A self-hosted voice agent platform: import your Dialogflow CX and CX Agent Studio agents, then build, test and take calls on your own hardware — for about $0.02–0.17 a call instead of $0.50.

• Drag-and-drop builder that mixes LLM steps and scripted pages, with a live test console
• 20 model providers (Claude, OpenAI, Gemini, Vertex, Bedrock, Ollama…) and phone/SIP via jambonz
• Agents as code (YAML + Markdown) with the vk CLI · local-first, no telemetry · built on Pipecat

🛣️ ioniq6-pilot

Personal openpilot build for a Hyundai Ioniq 6 on a comma 3X — NNLC steering tune, reverse-camera view, sentry mode and a phone dashboard, auto-rebased onto each new sunnypilot release. Python

🤝 HealthAssist — Production

Bilingual, cloud-native Dialogflow CX virtual agent on Genesys CX (us-central1/us-east1 redundancy) automating medical billing & records self-service. Deflects 20–40% of calls, ~85% lower contact-center cost, 24/7 availability.

🩺 Electronic Visit Verification Bot — Production

Multilingual caregiver visit verification over calls + SMS using Voximplant & Telnyx — real-time compliance tracking and streamlined reporting.

🤖 GenAI Suite — Production

GPT-4 + Flask chatbots and data-analysis tools that cut response times and turn large datasets into decision-ready insights.

---

🏆 Where I've Shipped — clients & employers

HCA Healthcare · Parallon · CereCore · Cognizant · Verizon · University of Maryland Medical Systems · Devoted Guardians · Prince George's Community College · Elara Caring

🎓 Certifications

• Google Cloud — Professional Data Engineer
• Microsoft Azure — AI-102: Designing & Implementing an Azure AI Solution

---

📊 GitHub Stats

---

If you want to grab coffee ☕️ and talk conversational AI, GenAI, or self-driving cars — reach out!