Applied AI Engineer

··

Phone agents that answer calls and do the work: registering callers in natural conversation, confirming every detail before saving, and letting verified callers update their record. Built for production, with signed webhooks, per-call cost tracking and automated tests.

Retell AIClaudeFastAPIMongoDB
Patient Registration Voice Agent screenshot

Patient Registration Voice Agent

Solo Engineer · 2026

A phone line where Sarah, an AI intake coordinator, registers new patients, verifies returning ones and updates their details. Call it, talk to it, then look at the record it left behind on the staff dashboard.

  • Registers callers in natural conversation, spells names back, reads a full summary before saving and gives a member ID. Runs in 12 languages, and names in any script work.
  • Built to protect privacy: it never confirms that a record exists to an unverified caller, and a failed verification never says which detail was wrong. Duplicates are caught quietly and shown to staff.
  • A full REST API and a staff dashboard with call transcripts, logs and possible duplicates sit behind it, covered by 174 automated tests.

languages on one phone line

Retell AIClaudeTool CallingFastAPIMongoDBReactRailway
Knowledge System screenshot

Knowledge System

Solo Engineer · Oct 2026

An internal knowledge assistant that answers staff questions from a company's own documents, cites every answer, and says when it doesn't know.

  • Demo over more than 2,100 public documents of a bank and its regulator, mapped to 17 departments. Cited PDFs open at the exact page.
  • "I don't know" is its own state: what was searched, the closest sources, who would know. Every one feeds a gaps view of what was never written down.
  • Access is enforced inside the database search, and the model never writes database queries. A trace shows each search, source, delay and cost.

+documents searched, every answer cited

RAGKnowledge GraphNeo4jMCPGeminiFastAPIReact

Stu

Co-Founder · Lead Engineer · Aug 2025 — Present

A B2B2C agentic AI company in the grocery space. Stu is an agentic food concierge with persistent memory that plans meals against local brands, PKR pricing and regional cuisine, then assembles the grocery cart end to end.

  • Shipped the consumer app to open beta — live as a web app with 100+ verified users.
  • Embedded a Stu-powered agent into a live Shopify storefront that builds the shopper's cart in place, over the store's own catalogue.
  • Ran the production surface: rate limiting, concurrency control, capacity capping, staged usage allowances, structured event logging and continuous deployment across backend and frontend.

+verified users in open beta

Agent ArchitecturePersistent MemoryStructured OutputsRAGTool CallingFastAPIReactShopify APIGoogle Cloud Run

Robi

Solo Engineer · Open Source · 2025

A physical AI agent that sees you, looks at you, and talks back — built end to end across firmware, wireless transport, mobile app and cloud AI.

  • Closed the perception-to-actuation loop: on-device face detection drives a BLE-linked pan/tilt servo rig in 1° steps, keeping Robi's head aimed at the user.
  • Streamed live audio and camera video to the Gemini Live API over WebSocket, fronted by an animated eye display rendering gaze and expression.
  • Designed the ESP32 firmware and multi-supply servo power architecture alongside the Flutter client.

technology layers, one system

Multimodal AIGemini LiveGoogle ML KitESP32BLEFlutterPWM/Servo Control

Stack

Agentic AI

Agent ArchitectureCustom OrchestrationLangGraphTool CallingStructured OutputsDeterministic RoutingPersistent MemoryRAGBM25 RetrievalMCPVoice AgentsPrompt EngineeringLatency & Cost Optimisation

Models & Runtime

ClaudeGeminiGPTMistralQwenOpenAI-Compatible APIsllama.cppLocal Deployment

Languages & Fullstack

PythonTypeScriptDartSQLC/C++ReactNext.jsFlutterFastAPINode.jsPocketBaseMongoDB

Systems & Cloud

System ArchitectureGoogle Cloud RunCloudflareDockerLinuxMiddleware IntegrationIBM MQ

Embedded

ESP32BLEMQTTWiFiPWM/Servo ControlUART

Building something in the physical world?

Agriculture, energy, small-scale local AI compute, fishing, tourism, robotics, restaurants and food. If you're working on something in any of these, I'd love to hear about it. No pitch needed, just say hello.

Say hello →
Would love to Chat →