Headline

AI Engineer | Forward Deployed Engineer: production voice agents, agentic workflows & integrations, end to end | 500+ live interactions/day | Python · TypeScript · Pipecat · Vapi · LangGraph · FastAPI · n8n | Remote

About

I'm an AI Engineer who works the way a Forward Deployed Engineer does: sit with the customer, figure out what actually needs to work, build it, ship it, and own it in production. Right now that's at ClinicPro AI, where I build and operate production voice agents and integrations for a multi-tenant platform handling 500+ live interactions a day. The domain is healthcare today; the work — voice AI, agentic workflows, and integrating with whatever system the client already runs — transfers to any sector. Most of my work sits at a boundary that doesn't show up in demos: deciding when an agent should stop talking and hand the person to a human. Escalate too eagerly and you've rebuilt the phone queue you were replacing. Let it push too far and it answers something it had no business answering. I build that line into the workflow — confidence thresholds, high-risk intent routing, retry limits, human-approval interrupts — rather than a prompt the model can talk itself out of. What that looks like day to day: • Running demos and requirement calls directly with clients, then turning those into working agents and rollout plans • Building the voice stack: a self-hosted Pipecat server (FastAPI, Gemini Live speech-to-speech, Telnyx + WebRTC), Vapi and Twilio agents with SIP, DTMF and warm transfers, and the SaaS platform clients use to run them • Integrating systems with no public API — reverse-engineered 8+ third-party platforms via HAR analysis, plus direct FHIR integrations — alongside CRM (GoHighLevel) and n8n automations • Orchestrating multi-step agents with LangGraph: branching on intent, persisting state across turns, interrupting for human approval • Owning the evaluation layer: 340+ adversarial tool-chain tests and persona-driven multi-turn tests asserting both responses and tool-call arguments against live agents, plus the full Playwright E2E suite • Keeping things running when infra breaks, and leading the architecture audit and reliability roadmap Stack: Python · TypeScript · FastAPI · Pipecat · Vapi · Twilio · Telnyx · LangGraph · CrewAI · AutoGen · MCP · n8n · Supabase/Postgres · pgvector · GCP · Docker · Playwright · PyTorch · Hugging Face. Claude Code daily. Before this: NASA Space Apps 2023 winner (1st of 700+ teams) with a transformer LLM built from scratch in PyTorch; Smart India Hackathon winner; HackMIT 2024 (2nd) and 2023 (3rd); led 70+ developers at MIT-WPU Innovation Hub. Remote-first and always have been. Open to AI Engineer and Forward Deployed Engineer roles in any sector. mann.pandya.18@gmail.com · github.com/mannpandya1702

Location

Bengaluru

AI stack

claude code

claude

codex

elevenlabs

gemini

openai

Work experience

AI Engineer at ClinicPro.ai, Jan 2026 - Present · 11 mos, Remote, Forward-deployed AI engineer on a multi-tenant voice AI platform — 500+ live interactions a day across clinic tenants. I own the voice server, the customer-facing SaaS platform, and every API and third-party integration we ship, and I work directly with clients from demo to rollout. • Built the self-hosted Pipecat voice server (FastAPI, Gemini Live speech-to-speech, Telnyx + WebRTC transports; patched Pipecat fork) and the web platform clients use to configure and monitor their agents • Shipped production agents on Vapi and Twilio — SIP trunking, DTMF, warm transfers, multilingual, number provisioning across US/India/UAE • Integrated 8+ third-party systems with no public API by reverse-engineering their web clients via HAR analysis, plus direct FHIR integrations for booking, cancellation, rescheduling and intake • Orchestrate multi-step agent workflows with LangGraph: branching on intent, persisting state across turns, interrupting for human approval on low-confidence and high-risk intents • Built the agent evaluation layer: 340+ adversarial tool-chain tests and persona-driven multi-turn tests asserting both responses and tool-call arguments against live agents; own the full Playwright E2E suite • Led the platform architecture audit (agent prompts, Supabase schema and security, n8n + GoHighLevel automations, integration adapters, CI/CD, HIPAA) and the phased reliability roadmap now in execution • Built a cross-sub-account SMS bridge (n8n + GoHighLevel API) that kept client messaging live after a carrier campaign suspension • Deploy and operate services on GCP; diagnose failures across the LLM, telephony and API layers under live traffic • Tuned intent classification and entity extraction across booking flows — cut missed appointments 40% and clinic admin workload 60% • Run demos and requirement-gathering calls with clients; turn them into agent specs, integrations and rollout plans

Junior AI Engineer & Web — Developer at ReechX, Oct 2025 - Jan 2026 · 4 mos, Remote, • Design, develop, and deploy AI/ML models for automation, prediction, and data-driven insights. • Build and integrate AI algorithms into scalable web applications using React.js, Flask, or Django. • Handle data pre processing, cleaning, and visualization to support model training and evaluation. • Implement and optimize RESTful APIs ensuring seamless communication between AI modules and web systems. • Manage version control, documentation, and deployment pipelines, while staying updated on emerging AI and web technologies

Artificial Intelligence Engineer at NuMinds AI, Jun 2025 - Sep 2025 · 4 mos, Remote, • Collaborated on AI model development and data pre processing for automation projects. • Enhanced system performance by optimizing training data pipelines and tuning hyperparameters. • Contributed to research on computer vision and NLP model integration for product improvement.

Cyber Researcher Analyst at DeepCytes Cyber Labs (UK), Jan 2025 - Mar 2025 · 3 mos, Mumbai, Maharashtra, India, • Designed and deployed AI-driven systems for threat intelligence and cyber risk mitigation. • Developed automated tools for real-time threat detection using LLMs and data analytics. • Led research on emerging cybersecurity trends and built defense models for vulnerability prediction. • Contributed to an AI-based Learning Management System for cybersecurity training.

CSE Committee Member at Innovation Hub ( Student's Innovation Council MIT WPU), Aug 2023 - Dec 2024 · 1 yr 5 mos, • Lead a team of 70+ developers, providing mentorship and project direction. • Delivered custom software solutions for clients including Capgemini and the Indian Army, ensuring 100% on-time delivery. • Organized weekly sessions on Data Structures & Algorithms, improving members’ problem-solving skills by 50%. • Initiated and executed 5+ major software projects, enhancing productivity across the team.

CSE Team Member at Innovation Hub ( Student's Innovation Council MIT WPU), Oct 2022 - Dec 2024 · 2 yrs 3 mos, Pune, Maharashtra, India

Education

Bachelor of Technology, Computer Science, MIT World Peace University

Resonance Eduventures Limited

Public links

github: https://github.com/mannpandya1702