PxE Holographic Imaging

**Product and the concrete problem it solves.** PxE Holographic Imaging attacks a structural inefficiency in modern machine vision: today a system that needs color, infrared night-vision, and 3D depth typically bolts together three separate subsystems — an RGB camera, an IR/thermal sensor, and a time-of-flight or structured-light depth unit — each with its own optics, illumination, calibration, power draw, and cost. PxE's pitch is to collapse all of that into a single, compact module built on a *standard* CMOS image sensor. By replacing the conventional Bayer color filter with a static, transparent diffractive-optics element the company calls the **HoloCoder**, PxE captures the incoming light's wavefront as a "white-light hologram" and then computationally reconstructs a native, per-pixel-registered stack of red-green-blue color, near-infrared, and depth from one exposure. The commercial promise is a camera with one lens, one sensor, and one processing chip that sees in the dark, perceives 3D, and costs closer to a commodity 2D camera than to a multi-sensor rig — a genuinely differentiated position if the reconstruction quality holds up in the field.

**Core technology and how it actually works.** The physics is the differentiator. Conventional cameras treat light as rays and throw away roughly two-thirds of incoming photons at the Bayer filter; PxE instead treats light as waves. The HoloCoder is a passive, transparent diffractive layer placed over an off-the-shelf CMOS sensor that encodes spectral and phase information into an interference pattern (a hologram) on the sensor plane. A proprietary, **fully deterministic real-time image signal processor** then decodes that hologram back into aligned RGB + IR + depth channels. Two properties matter for diligence. First, because the color-separating filter is removed, far more light reaches the sensor — PxE publicly claims on the order of **2x signal-to-noise ratio** versus a Bayer camera, with press materials citing effective light-sensitivity gains of roughly 3–4x, which is precisely the axis that matters for night, dusk, and adverse-weather vision. Second, PxE stresses that its decoding is *deterministic* rather than a trained neural network — a design choice meant to avoid the calibration drift, hallucination, and edge-case brittleness that plague learning-based depth estimation, at the cost of heavy optical and algorithmic engineering. The architecture's reuse of standard CMOS sensors and existing foundry/packaging flows is a deliberate manufacturability play, lowering the barrier to integration versus exotic sensor substrates.

**Market, customers, and go-to-market.** PxE is a horizontal imaging-IP company chasing a very large surface area: it frames its addressable market as effectively "every camera," and names automotive (ADAS and in-cabin driver/occupant monitoring), security and surveillance, robotics and drones, industrial machine vision, and consumer devices such as smartphones and laptops. The realistic go-to-market for a small deep-tech imaging startup is not to manufacture cameras at scale itself but to license or co-develop the HoloCoder optics-plus-ISP stack into the sensor and module supply chain — winning design-ins with image-sensor vendors, Tier-1 automotive suppliers, and camera-module makers. That is a long, reference-driven, design-win-dependent sales motion with multi-year automotive qualification cycles, and it is the single biggest execution question surrounding the company. PxE's presence at **CES 2025**, where it debuted a next-generation Holographic RGB-IR-Depth camera and was named a **CES 2025 Innovation Awards honoree**, signals a deliberate push from lab demonstration toward customer-facing productization and partner recruitment.

**Traction, funding, and third-party validation.** PxE should be read as an **early-stage** company whose validation is qualitative more than financial. It emerged from stealth in 2023 having raised approximately **$6.5 million** in total — anchored by a **$5.4 million seed round** — and, crucially, its backers are strategic corporate venture arms rather than generalist seed funds: **KDT (Koch Disruptive Technologies)**, the investment arm of Koch Industries, and **M Ventures**, the corporate venture arm of Merck KGaA. Strategics of that caliber conducting technical diligence and taking a position is a meaningful, if not conclusive, external signal for a hardware-optics company. The CES 2025 Innovation Award adds independent third-party recognition of the technology's novelty. The honest calibration is that the publicly confirmable evidence stops at demonstrations, awards, and strategic seed capital: there are no disclosed production design wins, named paying customers, revenue figures, shipping volumes, or a disclosed follow-on round since 2023, and headcount is not publicly confirmed. Anyone underwriting PxE is underwriting differentiated optics IP and a credible team, not proven commercial traction.

**Founders and team background.** The founding team is the company's strongest asset and the reason the deep-optics claims are credible. Co-founder and CEO **Yoav Berlatzky (PhD)** and co-founder and CTO **Yanir Hainick (MSc)** both come out of Israel's precision-metrology and semiconductor-inspection world — with prior roles at **Nova** (semiconductor metrology) and, for Berlatzky, **Applied Materials** — exactly the domain where extracting information from light at the physical limit is the day job. Each founder is credited with roughly **40+ granted and pending patents** across their careers, and Hainick is a graduate of the Israeli military's elite **Talpiot** program with prior IDF service, giving the team both deep computational-optics expertise and a defense-adjacent pedigree. This is a founder profile well matched to the problem: holographic wavefront encoding and deterministic reconstruction is a physics-and-algorithms problem, and the team's background is in exactly that. The principal team-level open question is commercial and organizational depth — the public record does not document a large business-development, automotive-qualification, or manufacturing-scaling bench, which is what converting optics IP into design wins requires.

**Competitive dynamics.** PxE competes on multiple fronts, and its edge rests on the single-sensor RGB+IR+depth fusion rather than on beating any one incumbent at its own game. (1) Against **CMOS image-sensor incumbents** (Sony, OMNIVISION/onsemi, Samsung) it is not trying to displace the sensor but to sit on top of it — though those giants also ship RGB-IR and depth-capable sensors and could pursue similar computational approaches. (2) Against **dedicated depth cameras** (Intel RealSense-style structured light, ToF modules from Sony/pmd, Orbbec, Lucid) it argues for lower cost, smaller form factor, passive operation, and no active illumination. (3) Against **thermal/IR incumbents** (Teledyne FLIR, Lynred) it offers near-IR sensitivity fused with color and depth rather than standalone LWIR thermal — a different, complementary capability. (4) Its closest true analogs are other **computational/meta-optics startups** — Metalenz (metasurface single-sensor imaging), Airy3D (transmissive diffraction-grating depth), and Israel's own TriEye (SWIR "see-in-the-dark" sensing) — each attacking the "richer data from a simpler camera" thesis from a different physical angle. PxE's plausible advantages are: the breadth of channels captured in one snapshot; the light-efficiency gain from removing the Bayer filter; the deterministic (non-ML) reconstruction; and manufacturability on standard CMOS. Its vulnerability is that meta/computational-optics is a hard, capital-intensive category where several well-funded players are racing for the same automotive and consumer design wins.

**Defense, security, and resilience dual-use relevance.** PxE's dual-use relevance is real but should be framed as a strong adjacency rather than a fielded defense capability. The core value proposition — passive, low-light color-plus-infrared-plus-depth perception from a single small, low-power module built on commodity sensors — maps directly onto several defense and homeland-security needs: SWaP-constrained ISR and situational-awareness payloads on small drones and ground robots; night and low-visibility machine vision for autonomous systems; perimeter, border, and force-protection cameras that need IR night vision and 3D ranging simultaneously; and counter-UAS or surveillance sensing where passive operation (no emitted illumination to detect) is operationally attractive. The Talpiot/IDF pedigree of the CTO and the security vertical PxE explicitly names reinforce that the company understands this market. The calibration that keeps this honest: PxE markets primarily to commercial customers (automotive, consumer, industrial, security), and there is no public evidence of fielded military programs, defense contracts, or qualified defense deployments. Dual-use here is a credible *enabling sensing technology* whose defense weight would rise sharply with any demonstrated robotics/ISR integration, not a demonstrated battlefield capability today.

**Growth stage, trajectory, and key diligence risks.** PxE reads as a differentiated but unproven **early-stage** deep-tech company: founded 2019, out of stealth 2023 on roughly $6.5M of strategic seed capital, shipping demonstrations and award-winning prototypes but not yet publicly documented production wins. The trajectory it wants — lab physics → CES-grade prototype → design-ins in automotive/robotics/security supply chains — is the right one, but the gap between an impressive single-snapshot demo and a qualified, high-yield, mass-manufacturable module is exactly where computational-imaging startups stall. The key diligence risks are, in order: (1) **commercialization and design-win risk** — no disclosed paying customers or production programs, and automotive qualification is a multi-year gauntlet; (2) **funding and runway risk** — the last disclosed raise was 2023 and modest for a hardware-optics company, so capital adequacy and dilution are open questions; (3) **manufacturing and yield risk** — diffractive HoloCoder layers must be produced at scale with tight tolerances and high yield to hit commodity-camera economics; (4) **reconstruction-quality risk** — the deterministic decoding must deliver depth accuracy, resolution, and robustness competitive with dedicated depth/IR systems across real-world lighting; (5) **competitive-timing risk** from better-capitalized meta-optics and sensor incumbents; and (6) **thin public disclosure** on headcount, revenue, IP ownership specific to PxE (versus founders' career patents), and any defense engagement. Progression from here would be evidenced by named design wins, a priced follow-on round, disclosed manufacturing partnerships, and independent benchmarks of depth/IR quality against incumbents.