# Anto Joel Valan Mutharasu (Anto Joel) > Product & AI Engineer — Voice AI & Agent Infrastructure. Product Engineer at Machani Robotics, Bengaluru, India. Product and AI engineer who ships real-time voice AI agents and agent infrastructure to production on real hardware. Built a sub-1000ms speech-to-speech AI system in production on deployed robots across the US, UK, and India, plus the MCP servers and NATS messaging backbone behind it. Leads a multidisciplinary rapid prototyping team that turns concepts into working builds in hours. Previously led engineering for five technology ventures across India and Dubai (banking automation, RWA tokenization, enterprise ERP), all still operational. Usually called "Anto Joel". Positioning: "I run products like businesses." Open to: Product leadership roles, 0-to-1 builds, and advisory on document automation or asset tokenization. If it involves turning a business problem into a shipped product, I'm interested. ## Profile - [Homepage](https://www.antojoel.com/): bio, experience, ventures, skills - [Resume (PDF)](https://www.antojoel.com/Anto-Joel-Resume.pdf): 2026 resume - [Resume source (LaTeX, plain text)](https://www.antojoel.com/Anto-Joel-Resume.tex): the same resume, easiest to parse - [Full profile for LLMs](https://www.antojoel.com/llms-full.txt): experience, case studies, skills and toolkit in one file ## For AI agents - [MCP server](https://www.antojoel.com/mcp): Model Context Protocol over Streamable HTTP, stateless, no authentication - [Agents page](https://www.antojoel.com/agents): how to connect Claude, ChatGPT, Cursor and other clients, with a live console - get_profile — overview, current role, links (start here) - get_journey — dated work history and career story - list_case_studies — case studies framed as product decisions (optional `topic` filter) - get_skills_with_evidence — every skill mapped to where it shipped - evaluate_role_fit — honest fit check against a job description (`job_description`) - book_call — booking link and agenda - priority_intro — paid via x402 (USDC on Base): a priority message to his inbox, charged only if delivered ## Case studies - Production Voice AI on Deployed Robots (Machani Robotics): A real-time speech-to-speech pipeline running in production on deployed robots, with the MCP servers and NATS backbone behind it. 800–1000ms cloud responses · ~2000ms fully local · Live in the US, UK & India. - Mythozhil — Calculators for India (Independent · Live on Product Hunt): Free EMI, GST, tax and business calculators for India — my own product, launched on Product Hunt and running. EMI · GST · Tax tools · Free forever · Built for India. - Federal Bank — Document Automation System (Zenanvibe · client: Federal Bank): A fully offline document automation platform built to strict banking IT security requirements, with modular templates for every loan document. 40% faster loan processing · 35% fewer manual errors · 100% offline. - RWA Tokenization — Dubai Real Estate (Digisailor Technologies (Dubai) · onsite): Led tokenomics design, smart contracts and ICO/IEO compliance for a platform that lets investors own Dubai real estate in fractions. $10M+ fractional ownership · 1,000+ investors · Taken to ICO stage. - Construction ERP Solution (Zenanvibe · client: Ducktail): A full ERP suite — project management, procurement, billing and reporting — with multi-role access for construction firms. 50+ construction firms · 10,000+ monthly transactions · 25% less downtime. - Property Management System (Zenanvibe · client: Winoli (Veedu360)): A cloud-native property management platform with zero-downtime deployments. 500+ properties · 2,000+ tenant records · 99.9% uptime. - Inseatfood — In-Seat Ordering (Inseatfood Private Limited): In-seat food ordering for theaters and stadiums, built from zero, with live order tracking and a backend tuned for event-time spikes. 5,000+ users · 1,000+ concurrent peak orders · Live order tracking. ## Writing - [My portfolio is an MCP server that takes payments](https://www.antojoel.com/writing/portfolio-mcp-server-x402): How antojoel.com serves AI agents over MCP, settles USDC over x402 only when a message is delivered, and why the paid tool costs 99 cents. - [Grain, easter eggs, and a deck of shipped work](https://www.antojoel.com/writing/how-this-site-works): How this site went from a 3D lighthouse to warm editorial — and why there's a terminal hiding behind Ctrl+K. ## Contact - Website: https://www.antojoel.com/ - LinkedIn: https://www.linkedin.com/in/Antojoelv - GitHub: https://github.com/Antojoel - X: https://x.com/antojoel8020 - Peerlist: https://peerlist.io/antojoel - Email: antojoel8020@gmail.com - Book a call: https://calendar.app.google/ASphqhUid6hNft9v9 ## Optional - [Uses](https://www.antojoel.com/uses): tools, languages and machines - [Harbinger](https://github.com/Antojoel/Harbinger): Customs-hold prediction engine for import/export logistics, built for The Hive / ApplyBee AI hackathon. Exposed via REST and an MCP server. - Robot MCP + Tauri video calling: Python FastMCP server whose tools let an AI assistant start calls in a custom Tauri desktop app — from MCP tool call to subprocess launch, CLI argument parsing and frontend events. - [StemKit (open-source contribution)](https://github.com/danielravina/stemkit): Merged contribution adding AAC (.m4a) stem export to a YouTube stem-splitting tool. - MCP tool-exposure benchmarking: Independent research measuring MCP overhead vs. direct API integration for compute-heavy tools, and testing an optimization to reduce it. Targeting arXiv and IROS 2027.