Vir Khanna
Forward Deployed Agent PM
Summary
AI Agent PM who still builds. My job is making agents that survive contact with real customers: I sit with them, learn how the business runs, and stay in the transcripts until the agent holds. I came to product from engineering, so when something bends I fix it myself.
Experience
Hello Patient — Healthcare AI
July 2026 – PresentForward Deployed Agent PMAustin, Texas
- I run customer accounts end to end: context sessions at the practice, then the build, the evals, and the launch onto live phone lines
- The agents are the real thing. They hold live voice conversations, reach into the systems of record mid-call, recover when a caller interrupts or changes their mind, and know when to pull in a human
- I audit production conversations, find where the agent bends, and ship the fix myself: call flows, prompts, code
- I help launch brand new customers, including the platform’s first practice on an EMR it had never touched
Nudge AI — Healthcare AI
October 2025 – July 2026Forward Deployed EngineerSan Francisco, California
- Built the AI voice receptionists that answer phones for healthcare clinics. Owned each from call flow design to deployment on real patient lines, tuning onsite until it held
- Shipped 15+ HIPAA-aware receptionists across multiple states. They book appointments, verify insurance, and route urgent calls with no hold
- Built an internal fleet of agents that took over marketing and sales work, saving about 6 engineering hours a week
Blendid AI — Food Robotics
June 2025 – October 2025Software Engineer, RoboticsSunnyvale, California
- Owned the launch of automated refund flows across 15 robotic kiosk sites, cutting ticket resolution latency 72%
Everise — BPO, Customer Experience
June 2024 – December 2024Software Engineer, Voice AIPlantation, Florida
- Rolled out voice AI agents across four global IT help desks, automating about 70% of inbound call resolutions
Education
University of California, Davis
September 2025B.S. Computational Cognitive Science, Minor in PhilosophyGPA: 3.8
- Founded the Machine Learning Student Network and grew it to 120+ students shipping their first production ML
Projects
Bolo — neural TTS engine, from-scratch MLX-Swift port
Swift, MLX-Swift, SwiftUI, Groq
- Ported the Chatterbox-Turbo neural TTS engine from Python/MLX into native MLX-Swift (T3, S3Gen, vocoder)
- Shipped a public MIT macOS menu-bar app: Groq read-aloud with system-wide hotkeys and Accessibility capture
- Validated the port with six numerical parity gates; tuned for Apple Silicon via 4-bit quantization, KV cache
Brain — persistent memory pool for stateless AI agents
Agent Skills, MCP, Obsidian, Telegram Bot
- Engineered a 3-layer memory pool (capture, curated wiki, schema contract) so agents resume with full context
- Automated a Telegram-to-vault pipeline (VPS, git, a 5-skill triage, entity-resolved wiki) agents query live
Skills
- Languages
- Python, TypeScript, Swift, SQL, Bash
- ML & AI
- LLM apps, AI agents, prompt engineering, RAG, evals, voice and conversational AI, NLP
- Agents & Automation
- LangGraph, LangChain, CrewAI, MCP, n8n, Zapier, Jinja, LangSmith
- Voice Stack
- LiveKit, ElevenLabs, Retell AI, Twilio, Deepgram
- Technologies
- React, Next.js, Django, FastAPI, Node.js, PostgreSQL, Redis, WebSockets
- Tools & Infra
- Git, Docker, AWS, GCP, REST APIs, CI/CD
- Product
- A/B Testing, Product Roadmapping, Telemetry, UX, PRDs, Technical Writing