Aung Myint Oo

I'm the founding engineer and technical lead at Salesbugle, an AI sales coaching SaaS I built from an empty repository to production with paying customers, as the sole engineer. Before that I spent two years as a robotics and vision engineer at Mozark, building robotic mobile-device testing systems that run in four countries. I hold an MSc in Robotics from the National University of Singapore and a B.Eng in Electrical Engineering with a specialisation in robotics.

What I do

I take AI products from idea to production, end to end: scoping, architecture, the AI systems, infrastructure, launch, and the operations that keep a product alive afterwards. I work with founders and domain experts who know their industry deeply and need one person to own everything technical, as a founding engineer, a contract technical co-founder, or a long-term, equity-aligned technical partner.

Robotics and perception

At Mozark I led the robotics and computer vision work behind a robotic mobile-app testing platform: physical Delta robots interacting with real phones, driven by vision rather than device hooks, for fintech apps where native automation frameworks are blocked by security controls.

• Delta-robot automation framework for mobile-device interaction, including G-code generation and a scale-and-shift coordinate system; I also designed and fabricated the device mounts and casings.
• Computer vision and OCR pipelines for UI detection and visual assertions: OpenCV, SIFT/FLANN template matching, image segmentation, edge-detection auto-deskew, video quality analysis, and ML UI detection with Omniparser.
• Video processing for latency KPIs (ffmpeg, GStreamer), asynchronous pipelines at sub-100 ms latency.
• Deployment, maintenance, and repair of the systems across Singapore, Philippines, Thailand, and India.
• Optics Framework, an open-source no-code test automation framework on PyPI that unifies Appium, vision-based matching, and the robot backend, with an agentic LLM layer that turns natural-language instructions into executable tests.

My NUS MSc thesis extended the same stack into a three-stage pipeline: natural language → computer vision → robotic motion, using ROS and OpenCV, so a robot can act on a phone screen from a plain-English instruction. Coursework included Robot Perception, Deep Learning for Robotics, Autonomous Robot Systems, Human-Robot Interaction, and Fuzzy/Neural Systems for Intelligent Robotics. As an undergraduate I trained a 2-DOF pick-and-place arm with deep Q-learning, built an autonomous facade-cleaning robot, and built a VR haptic glove with its own Unity environment.

Voice, vision, and language in one system

Perception is more than one modality. At Salesbugle I built a real-time meeting AI that joins live video calls as a participant, presents a deck as its camera tile, listens to the room through a streaming speech pipeline (VAD, sliding-window Whisper transcription, speaker diarization, multi-layer hallucination filtering), and answers grounded questions when addressed by name. Alongside it sit agentic chat systems that call tools over live business data, propose writes, and execute them only after confirmation, with a full audit trail. The interesting systems combine vision, audio, and language with action.

Learning systems that compound over time

The question I keep building around is how a system gets better from its own experience instead of starting from zero every session. amoOS is my working answer: a personal AI operating system where a fleet of AI coding agents does the building across eleven projects while I plan, dispatch, and verify. Every session, decision, and lesson is distilled into a compiled knowledge base, currently over four hundred pages, that the agents load as context for the next task. A lesson learned in one product immediately benefits the next, and the system measures its own utilization, yield, and attention cost. At Salesbugle the same idea runs in production as layered memory (semantic, episodic, procedural) over vector and relational storage.

To be precise about what this is and isn't: it is knowledge and memory infrastructure that makes agents measurably more capable with every task, with a human verifying every result. It is not online policy learning or reinforcement learning from physical experience. Closing that same loop on physical systems, robots that improve from their own operating experience, is the problem I most want to work on next.

AI agents in production

• Salesbugle: multi-tenant FastAPI and PostgreSQL backend on AWS ECS, LLM pipelines for sales-conversation intelligence, model-agnostic orchestration, and sole ownership of CI/CD, observability, cost tracking, and production operations.
• amoOS MCP: a hosted remote Model Context Protocol server exposing the whole factory as 86 tools to any MCP-capable agent (Claude Code, Codex, or anything that speaks the protocol).
• A climate-tech intelligence platform MVP shipped solo: sensor ingestion, time-series processing, anomaly detection, and an analytics dashboard.

Background

Founding Engineer & Technical LeadSalesbugle, Singapore · Oct 2025 - present

AI sales coaching SaaS, built from an empty repository to production with paying customers as the sole engineer.

Founding AI Architect (contract)Climate-tech intelligence platform, United States · 2026 - present

Architecture and end-to-end MVP: sensor ingestion, time-series processing, anomaly detection, analytics. Shipped solo.

Robotics & Vision EngineerMozark, Singapore · Aug 2023 - Sep 2025

Led development of robotic mobile-device testing systems deployed in Singapore, Philippines, Thailand, and India.

MSc Robotics (part-time)National University of Singapore · 2023 - 2026

Thesis: natural-language control of a robot interacting with mobile devices through computer vision.

B.Eng Electrical Engineering (Honours)Specialisation in Robotics; second major in Innovation & Design · 2019 - 2023

Projects included a 2-DOF pick-and-place arm trained with deep Q-learning, an autonomous facade-cleaning robot, and a VR haptic glove.

Based in Singapore

I live and work in Singapore, and work with founders in Singapore, the United States, and remotely. The systems I have shipped run in Singapore, the Philippines, Thailand, India, and the US.

Not to be confused with

Several people share the name Aung Myint Oo, including a general surgeon in Singapore and public figures in Myanmar. This site, github.com/davidamo9, and linkedin.com/in/aung-myint-oo99 all refer to the engineer described on this page. I also go by AMO.

Get in touch

Sitting on an idea you don't know how to build, a workflow that needs a modern solution, or a robotics or AI problem you want a second pair of hands on? The first conversation is free.

Last updated 2026-08-30.