Skip to content

Case notes

What I've actually built and shipped.

Selected proof points, from semiconductor inspection and enterprise computer vision to the agentic systems and audio experiments running today.

Guardian Agents: Founder (Dec 2025-present)

Context. Founded December 2025: a Physical AI / AgeTech product for healthcare operations.

What I did. Founder. Set the product direction across operational visibility and intelligence, quality control, and auditable risk mitigation. I am building it 0-to-1.

Outcome. Early-stage and building. This is the same founder-stage product work I bring to fractional engagements, done firsthand and in parallel.

Applied Materials AERA: closing the capability gap that mattered

Context. AERA was Applied Materials' next-generation aerial-image mask-inspection platform. Competitive research exposed a strategic gap against KLA: phase-shifting chrome-and-glass masks were effectively transparent to our existing inspection path. Separately, polarization control was a prerequisite for Intel acceptance.

What I owned. I carried both capability projects from market and customer evidence through feasibility, management approval, budget, optics, mechanics and opto-mechanics planning, subcontractor selection and coordination, manufacturing, QA, and acceptance. Both workstreams relied on specialist partners in electro-optics, micro-electronic glass etching, and delicate, high-accuracy mechanics.

The hard part. The off-the-shelf wave plates sold as zero order did not behave correctly across AERA's bundle of optical angles. I developed a new method to measure effective retardance, thickness, and polarization purity, escalated the evidence to Thorlabs' CTO, and gave the supplier a repeatable way to produce and verify true zero-order parts.

Outcome. The phase-contrast demonstrator and plug-and-play linear and circular polarization module were delivered on time, on budget, and to specification under an aggressive timeline. Intel accepted the polarization capability and purchased two AERA systems, opening a strategic adoption path for the platform.

AERA case map

Two capability projects. One commercial and technical loop.

This was not a handoff from market research to engineering. It was one connected product decision carried through physics, suppliers, manufacturing, acceptance, and adoption.

  1. Discover

    • Market
    • Customer
    • Competition

    Find the purchase gap

    Market and competitive research showed that KLA could inspect phase-shifting glass masks while AERA could not. Intel also required polarization control.

  2. Understand

    • Product
    • Physics
    • Technology

    Trace the physics

    The problem crossed phase contrast, chrome-on-glass features, polarized stepper conditions, wave-plate behavior, and a converging bundle of optical angles.

  3. Evaluate

    • Architecture
    • Evidence
    • Budget

    Select the evidence path

    I compared phase-contrast approaches, selected a feasible concept, won management budget, and defined how the optics and polarization purity would be measured.

  4. Implement

    • R&D
    • Suppliers
    • Manufacturing
    • QA

    Build and manufacture

    I delivered a phase-contrast demonstrator and plug-and-play polarization module across optics, precision mechanics, opto-mechanics, etched glass, suppliers, and acceptance testing.

  5. Optimize

    • Acceptance
    • Adoption
    • Supplier quality

    Feed evidence forward

    My measurement method exposed wave plates sold as zero order that did not behave as zero order. Corrected parts passed Intel acceptance—and the evidence improved the module, the supplier process, and the next-customer story.

One operating envelope

  • Competitive research
  • Phase contrast
  • Polarization optics
  • Precision mechanics
  • Supplier development
  • Customer acceptance

BriefCam: Director of Product (2018-2022)

Context. An enterprise video-analytics company running an AI/ML R&D organization of 30+ PhD algorithm developers building computer-vision and deep-learning algorithms.

What I did.
  • Took the license-plate-recognition (LPR) feature from concept to ship, including a camera-agnostic specification and a fuzzy, Levenshtein-distance plate search built for real-world cases like Amber Alerts. Exact-match search misses partial, misread, or transposed plates; fuzzy search finds the vehicle an operator is actually looking for.
  • Re-architected the video-analytics pipeline from centralized to distributed processing, deciding per algorithm what ran on the camera versus the central server, including a two-stage face pipeline (detection at the edge, recognition in the core).
  • Instituted a QA methodology for AI features: algorithm owners define their own pass/fail criteria and test plans before handoff, because they're the only ones who know where a model's envelope actually breaks.
  • Authored the hardware-sizing methodology (GPU/CPU benchmarking, storage and chassis right-sizing) that became the sales engineering standard.
  • Led the product organization: three senior PMs, two UX designers, and a technical writer.

Outcome. Enterprise deployments scaled from roughly 40 to 1,200+ cameras per installation, alongside a ~30% reduction in hardware package costs that kept upgrade pricing flat as the algorithms grew more demanding.

Mindful Designs Haven: Founder & Creative Director (Sep 2023-present)

Context. A generative-AI print-on-demand studio.

What I did. End-to-end, solo: AI image-generation for product design, multi-platform storefronts (Amazon Merch by Amazon, Etsy, Redbubble, WooCommerce/Elementor, Shopify), and performance marketing across Meta, Google, and Amazon ads.

Outcome. A live, AI-native commerce operation I've run continuously since 2023: the practical, hands-on counterpart to the enterprise AI work above.

Explore the print-on-demand studio

The agentic workspace: personal practice (ongoing)

Context. My own production environment: not a client project, but the infrastructure I use to run everything else.

What I did. Designed, built, and operate a full-stack agentic AI system: 12+ autonomous agents coordinated across research, outreach, business development, product operations, analytics, CRM, and DevOps. Underneath it: a graph-backed knowledge layer, multi-model LLM orchestration across providers, and an access-control architecture where each agent only sees the data its task requires.

Outcome. I'm the sole operator. When an orchestration failure or a data-flow fault happens, I debug it myself, in production. That's the operating discipline behind the consulting work.

Adaptive noise masking - live web experiment

Context. A browser experiment that listens to ambient sound locally, then shapes white or pink noise around the room's spectrum.

What I did. Framed the product question, built the interactive web proof, kept microphone analysis on-device, and made the core comparison testable on a phone.

Try it. The experiment is public now. Headphones are optional; the most useful comparison is in the room you want to mask.

Open the Noise lab Explore the audio practice

Want to talk about what you're building?

Start a conversation More about my path →