Robotics software + validation engineer
Software engineer and testing leader with 10+ years across XR, web, native platforms, autonomous systems, and system-level validation. M.Eng. in Robotics & Intelligent Autonomous Systems, focused on ROS 2, simulation, controls validation, automation, and CI/CD quality gates.
Featured robotics validation system
A distributed ROS 2 warehouse-control POC coordinating four AGVs, 24 cases/SKUs, rack storage, traffic lanes, chargers, conveyors, and two picker/palletizing stations. Its qualification suite probes uncertain outcomes, concurrent reservations, event delivery, process recovery, fleet scaling, and deployment behavior.
Main body of work
ROS 2 / Gazebo / SQA Automation
Distributed warehouse orchestration with AGV routing, DDS network faults, idempotent retries, crash recovery, event deduplication, fleet scaling, Gazebo, and Kubernetes deployment checks.
Open full case studyJetson Orin / OpenCV / xArm
OpenCV vision gate on an Orin Nano triggers a physical xArm cube grab through Hiwonder controller commands, readback checks, and trial logging.
Open full case studyROS Gazebo / OpenCV / PID Control
Simulated vehicle steered by a ROS Python PID controller using HSV masking, Canny edges, Hough lines, and vanishing-point error from a camera feed.
Open full case studyUnity ML-Agents / PPO / Raycasts
Quadcopter simulation trained with PPO, curriculum learning, behavioral cloning, GAIL, raycast observations, and randomized obstacle spawning.
Open full case studyUnity ML-Agents / Zero-G / Imitation Learning
Zero-gravity servicing-robot simulation trained with PPO, BC, and GAIL to rendezvous with a moving satellite and perform a soft-touch end-effector contact.
Open full case studyProfessional edge
ROS 2/ROS, simulation, robot kinematics, PID validation, control verification, and autonomous workflows.
Functional, integration, system, and performance testing for hardware-software and real-time systems.
Automation frameworks, GitLab CI/CD quality gates, Docker workflows, release validation, and traceability.
Meta Quest, Unity VR, therapeutic branching flows, voice recognition, device coverage, and performance analysis.
Web, native, API, database, and cross-platform validation across SQL, MongoDB, and Azure Cosmos DB.
JIRA, TestRail, root cause analysis, KPI metrics, defect lifecycle ownership, and release sign-off.
Web QA automation + LLM-guided exploration
Separate from the robotics work, this project shows modern web testing depth across browser behavior, API contracts, accessibility, performance, security, dependency checks, pytest, CI artifacts, and bounded multimodal review through the OpenAI Responses API.
Most recent run / generated report hub
The suite produces a polished report hub linking Playwright, Lighthouse, pytest, npm audit, optional ZAP, and AI-assisted artifacts. A blind GPT-5.6 review of desktop and mobile evidence returned an advisory verdict with three findings, each tied to evidence and a deterministic follow-up.
Browser-visible evidence of the Playwright, Lighthouse, pytest, npm audit, and cross-tool report artifacts generated by the suite.
Blind multimodal evidence review, structured bug decision, confidence, uncertainty, and Playwright-ready follow-ups in one visible workflow.
Navigation, filters, auth, task CRUD, form validation, desktop, mobile, traces, screenshots, and video.
Playwright request tests plus pytest/jsonschema checks for health, auth, CRUD, error shapes, and latency.
axe-core WCAG A/AA scans, keyboard entry, skip link behavior, dialog focus containment, and escape recovery.
Browser timing budgets, resource-weight budgets, Lighthouse scores, FCP/LCP/TBT/CLS, and report artifacts.
Headers, authz gates, logout invalidation, rate limiting, oversized input handling, npm audit, and optional ZAP.
GitHub Actions uploads reports, traces, media, and optional ZAP output, with advisory AI triage available after deterministic failures.
Desktop and Pixel 7 screenshots are paired with semantic and runtime context, reviewed without a seeded answer, and returned through a strict schema. Findings stay advisory until reproduced by an exact test or accepted through a reviewed threshold.
Executed small-model qualification / RAG / agents
A runnable CPU proof of concept uses the Apache-2.0 SmolLM2-135M-Instruct model. The retained red release gate demonstrates failure localization across data, retrieval, generation, and agent policy.
Retrieval passed 4/4 cases, but the model missed grounding and tool-use contracts. Invalid agent output was denied before any side effect.
Inspect the executed PoCValidate retrieval separately from generation so an unsupported answer can be localized to the corpus, ingestion, search, context assembly, or model.
Treat data defects as first-class product defects, with contracts and reconciliation at each transformation before data reaches retrieval or inference.
Replace brittle string equality with invariants, repeated trials, calibrated scoring, and regression tolerances while keeping critical failures deterministic.
A correct final answer is insufficient. Assert invariants across the plan, every tool boundary, state transition, side effect, and termination path.
Unit checks, schemas, permissions, prompt contracts, mocked tools, and a fast golden subset.
Full evaluation corpus, retrieval metrics, quality slices, drift checks, cost, and baseline deltas.
Adversarial scenarios, bias analysis, human review, lineage, approvals, and reproducible evidence.
Quality SLOs, sampled trajectories, drift and failure clusters, user feedback, and new regression cases.
XR testing in health and wellness
Professional Quest and Unity testing across health and wellness VR products where test quality protects therapeutic flow, session data cleanliness, user comfort, and release confidence.
Lead Test Architect + Test Engineer
I led test strategy and hands-on testing for therapeutic and wellness VR experiences spanning branching paths, scripted interventions, voice-recognition flows, Quest device behavior, session persistence, clinician-facing data expectations, and performance budgets.
CBT-based VR program testing for guided reflection, values, goals, and multi-week therapeutic progression.
Chronic-pain VR testing across movement, calming practices, patient safety, and session reporting.
XRHealth describes First Resort as an 8-week CBT-based VR program for emotional well-being, patient reflection, values, goals, and thoughts-to-actions education.
Product guideXRHealth frames Mind & Motion around chronic-pain support through gentle movement, calming practices, and patient education about how pain works in the brain and body.
Product guideMORE-VR applied mindfulness-oriented recovery enhancement in VR as an adjunct to medications for opioid use disorder, with session data, craving/affect measures, and safety considerations.
Phase 1 papergameChange uses CBT-style VR interventions and a virtual therapist to help people with psychosis practice everyday social situations at their own pace.
Study overviewBranch-path matrices, intervention scripts, state transitions, onboarding, session resumption, failure recovery, and program progression rules.
Voice-recognition prompts, response handling, fallback paths, controller input, gaze selection, comfort boundaries, and timed therapeutic events.
Meta Quest Scriptable Testing Services for repeatable device setup, AutoDriver recordings and playback for E2E input coverage, XR Simulator test runs, Unity Test Framework coverage, NUnit assertions, unit tests, integration tests, and release regression suites.
Validation focused on patient assignment, session metadata, outcome capture, clean report generation, and preserving analyzable data for therapeutic efficacy work.
Full manual test-suite definition for headset setup, first-run onboarding, accessibility, comfort, content progression, edge cases, retries, and cross-device release checks.
OVR Metrics Tool passes using HUD and CSV report evidence for FPS, stale frames, screen tears, heat, throttling, CPU/GPU, memory, app GPU time, and custom metrics, plus security-minded testing around permissions, account/session behavior, platform services, and privacy-sensitive workflows.
About
Robotics-focused engineer with a professional background in software engineering, SDET work, QA leadership, and system-level validation across XR, web, native, and simulation environments. My current emphasis is building robotics software and the automated evidence that proves it works.
ROS 2/ROS, IsaacSim, OpenCV, kinematics, PID, autonomous systems, control validation.
STLC, functional/integration/system/performance testing, JIRA, TestRail, RCA, KPI metrics.
Playwright, Cypress, Selenium, NUnit, GitLab CI/CD, Docker, Python, TypeScript, C#, C++.
Meta Quest, Unity VR, therapeutic XR testing, web/native QA, APIs, SQL, MongoDB, Azure Cosmos DB.