The product
An AI that answers the business line. Not a phone tree. Callers interrupt, ramble, and change their minds. HandlerAI keeps up, grounded on that customer’s knowledge base.
Founder of HandlerAI — a 24/7 AI phone receptionist for local businesses. I design the product, the cloud, and the full stack, and I run the company alone with AI.
I founded HandlerAI and run it as a one-person company — product, infrastructure, voice, billing, and operations — with AI as the rest of the team.
Multi-disciplinary technologist with 5–8 years spanning systems architecture, infrastructure, full-stack development, and team leadership. I use that range every day to ship HandlerAI without a staff.
Azure-certified cloud architect. Production AI across realtime voice, RAG, and tool-calling agents. Proven ability to operate as a one-person technical department — now applied to a company I own.
Founded 2026 · Hamilton, Ontario · Solo founder
HandlerAI is a 24/7 AI phone receptionist for local businesses. It answers in the owner’s name, knows their services, captures the lead, and can book the job — then texts them in seconds. I founded it. I still run every layer of it myself.
An AI that answers the business line. Not a phone tree. Callers interrupt, ramble, and change their minds. HandlerAI keeps up, grounded on that customer’s knowledge base.
Trades, clinics, salons, restaurants, auto shops — businesses where the phone is the sales channel and the person who should answer is already doing the work.
Multi-tenant SaaS. Isolated knowledge, numbers, and leads per customer. I designed the architecture so one operator can run many businesses without a call center.
HandlerAI is not a single model behind a prompt. It is several AI systems with different jobs — conversation, retrieval, tools, and operations — wired so a missed call becomes a lead.
Inbound audio streams from Twilio into a custom Voice Bridge on ECS Fargate. ElevenLabs Convai is the primary conversation model (persona Maya, voice Amber). xAI Grok Realtime is the reserve path. Opening greetings are pre-rendered TTS. Office ambience is mixed under the voice so it sounds like a desk, not a bot.
Each tenant gets an isolated Amazon Bedrock Knowledge Base with Titan v2 embeddings. During a call the agent can lookup_knowledge — hours, services, pricing, FAQs — so answers come from that business, not a generic script. A shared guardrail sits on generation.
The voice agent has tools: save a lead, look up a customer, book or cancel an appointment, end the call. Tools are gated by vertical and subscription tier. After the call, a fallback extractor still captures a lead if the model did not.
Onboarding is an AI-assisted interview, not a settings dump. Provisioning is an orchestrated ledger: knowledge base, ElevenLabs agent, Twilio number. An AI-ops agent (Bedrock Nova) labels, diagnoses, and proposes fixes from structured logs — so I can operate production without a NOC.
Cloud, telephony, data, and product surfaces designed so one founder can ship and operate a multi-tenant voice platform.
Caller dials a Twilio DID. Signature-validated webhook. Demo line has a press-1 robocall gate.
Lambda returns TwiML. Media Streams WebSocket to ECS Fargate. Bidirectional µ-law audio to ElevenLabs or Grok.
Knowledge lookup, lead save, booking — scoped to that customer, that tier.
Leads and bookings land in Supabase with org isolation. Owner gets SMS and email. Stripe meters the call.
AWS us-east-1. Numbered Terraform stacks with separate state and IAM — stacks talk only through SSM, not remote state. Lambdas for telephony, leads, booking, Stripe, onboarding. Voice Bridge on ECS Fargate behind an NLB. Secrets Manager for keys. CloudWatch to Stripe for usage.
Marketing site at handlerai.app. Authenticated customer portal and operator Mission Control as a separate app bundle on the same CloudFront distribution. Node APIs on Lambda Function URLs. Supabase (Postgres, Auth, RLS) as CRM. Stripe Checkout and metered overage. React/Vite on the web; the original mobile direction is Flutter.
Currently: Founder of HandlerAI. Also operating at DIT.
Operating across five distinct roles: systems architect, system administrator, Flutter developer, front-end web developer, and IT team lead. Responsible for the full technology lifecycle from strategic architecture decisions through hands-on implementation and team management.
Approach
I believe the best technology is invisible — it simply works. HandlerAI is that idea as a company: the phone gets answered, the lead is captured, and the owner keeps working. I built the systems so I could run it alone.
Daily driver for HandlerAI — architecture, code generation, review, and documentation. The reason one person can ship a voice platform, a billing stack, and a company.
Expert-level use for rapid prototyping, multi-modal processing, and leveraging Google's AI ecosystem for analysis, content generation, and workflow automation.
Designed and built retrieval-augmented generation systems connecting LLMs to structured and unstructured internal data sources using vector databases and embedding models for context-aware AI responses.
Hands-on experience evaluating, integrating, and deploying multiple large language models in production environments. Skilled in prompt engineering, model selection, cost optimization, and building AI-powered features.
Building with AI, hiring for a systems role, or want HandlerAI on your line? Let’s talk.
sandhudaman0024@gmail.com