K·P·C
Personal platform recordSheet 1 of 11

Kameshwaran Panchagnula Chandrasekar

Automotive Edge Architect & Systems Engineering Leader

BrandThe IndustrializerTakes emerging compute to production-ready, safe, manufacturable reality at scale.

See the recordDrive the journey
BOXABSoCHPCHPCZNLZNLGEN 1'011 box1 screenGEN 2'132 boxes4 screensGEN 3'191 SoC6 screensGEN 4'252 HPC + 2 zonalwhole vehicle
HPCCompute boxZonalDisplayS/A node
Vehicles on the road UNITS
10M+
Lifetime production volume. Platforms that shipped, in cars people bought, not a forecast.
Years shipping compute
25+
YEARS
Markets reached
6
REGIONS
Production platforms
8
PROGRAMS
02Impact · Markets reached

From regional launch to worldwide. Roughly ten million vehicles.

Each figure below is a lifetime worldwide production volume, not a forecast, and not split by country. These are platforms that shipped, in cars people bought.

  • Holden iQ

    Australia

    First production program, media playback in the cabin.

    150Klifetime units

  • Renault MM2012

    Europe

    Navigation and infotainment at mass-market volume.

    3Mlifetime units

  • Hyundai Genesis DIS 2.0

    Worldwide

    Premium cockpit delivered via LG.

    30Klifetime units

  • JLR InControl Touch Pro

    Worldwide

    Premium infotainment across the JLR line.

    2.1Mlifetime units

  • Chery Exeed LX, Hongqi HS5/HS7

    China

    Workload consolidation onto a single SoC.

    5Mlifetime units

  • JLR SDV platform

    Current

    Worldwide

    Software-defined vehicle on HPC and zonal compute.

    Volume not disclosed
03Selected work · How the problem was solved

The parts list says what shipped. This says how.

Platform work is judged on the decisions, not the deliverable. Each of these is written the same way: what was hard, what could not move, what was done, and what it produced.

Workload consolidation

The first x86 virtualization solution in a production cabin.

Era
2019-2020
Silicon
Intel Apollolake · ACRN
Market
China
Units
5M

Problem

No reference blueprint existed. Nobody had shown concurrent rendering across multiple isolated virtual machines at 60fps, and no automotive bootloader could bring up several operating systems on a single x86 SoC. The pieces existed separately. The proof that they could work together did not.

Constraint

Mixed criticality on shared silicon. The safety domain cannot be disturbed by Android infotainment, so freedom from interference had to hold under real load, not just on paper.

Approach

Prove it on two hypervisors, both on Intel x86. A QNX Hypervisor stack runs QNX RTOS on the safety domain and Android on infotainment. An ACRN stack adds a Linux Service VM to broker devices, with Zephyr OS on safety and Android on infotainment. Each brings up a safety RTOS and Android together from a purpose-built automotive bootloader, sharing one GPU so every display renders concurrently at 60fps rather than time-slicing.

Outcome

The first x86-based virtualization solution in automotive, and the reference that later programs were built against. The ACRN stack shipped across three vehicle programs at two Chinese OEMs: Chery Exeed LX in 2019, then Hongqi HS5 and HS7 in 2020.

QNX HYPERVISOR STACKCLUSTERIVIQNX RTOSSAFETYANDROIDIVIQNX HYPERVISORINTEL x86 · 1 SoCshared GPU · 60 fps concurrentACRN STACKCLUSTERIVIZEPHYR OSSAFETYANDROIDIVILINUXSERVICE VMACRNINTEL x86 · 1 SoCshared GPU · 60 fps concurrentFreedom from interference on both stacks

Fig 3.1 · Two x86 hypervisor stacks on one SoC

  • 2x86 hypervisor stacks
  • 60fps concurrent, per display
  • 3vehicle programs
  • 5Mlifetime units

Premium infotainment

A ten-fold faster boot, and a rearview camera that met its federal deadline.

Era
2016-2023
Silicon
Intel Bay Trail
Market
United Kingdom
Units
2.1M

Problem

JLR's flagship infotainment platform could not clear its ok-to-ship gate. Full-system start-up ran about ten times over budget, the rearview camera could not appear inside the deadline that U.S. federal law sets for it, and the IVI and rear-seat displays could black out. Any one of those was enough to block production launch.

Constraint

A rearview camera is a safety function, so the time it takes to appear is not a preference, it is a federal requirement (FMVSS). And the launch decision was business critical: every JLR vehicle line was waiting on this one platform.

Approach

Treat start-up as an architecture problem, not a tuning problem. I designed the boot sequence, the start-up sequence the system runs from power-on, so the rearview-camera image met the FMVSS deadline, and built a new software-based start-up strategy that brought full-system start-up down to about twelve seconds, roughly a ten-fold reduction. I designed novel mechanisms to make start-up and shutdown deterministic, so the platform came up and powered down predictably every time, and used FMEA to find and remove the root causes of the IVI and rear-seat black screens.

Outcome

The work cleared the barrier to production launch and supported the business-critical ok-to-ship decision. InControl Touch Pro rolled out across every JLR vehicle line, reaching roughly 2.1 million customer vehicles from 2016 to 2023, with the black-screen occurrences eliminated.

FULL-SYSTEM START-UP · SHARED TIME SCALEPRE-ARCHITECTURE · BLOCKED OK-TO-SHIP10x FASTERRE-ARCHITECTED · 12s FULL SYSTEM012s60s120sRearview camera meets the FMVSS deadline · deterministic start-up and shutdown

Fig 3.2 · Full-system start-up, before and after

  • 12sfull-system start-up
  • 10xfaster boot
  • 2.1Mvehicles, 2016-2023
  • 0black-screen faults
04Trajectory · Career record

Five places, one continuous thread.

  1. 2001-2013

    India

    Mistral, then Harman Connected Services

  2. 2013-2016

    United Kingdom

    Harman Connected Services

  3. 2016-2017

    Ireland (Shannon)

    Intel

  4. 2017-May 2023

    Chandler, Arizona

    Intel Corporation

  5. May 2023-Q3 2025

    Chandler, Arizona

    Jaguar Land Rover

  6. Q4 2025-present

    Portland, Oregon

    Jaguar Land Rover

05The arc · Two decades of consolidation

One screen in 2001. The whole vehicle by 2025.

Plotted on real time: every step up the ladder of complexity drove more surfaces from fewer, more capable boxes.

36911GEN 1200124GEN 2201316GEN 3201949GEN 42025
  1. GEN 12001

    Consumer Multimedia

    ADI Blackfin

    One screen in the cabin

    1 surfaces1 boxes

  2. GEN 22013

    Cockpit SoCs

    Intel Bay Trail

    Cluster, IVI and rear seat

    4 surfaces2 boxes

  3. GEN 32019

    Workload Consolidation

    Intel Apollolake, ACRN

    Many OSs on a single SoC

    6 surfaces1 boxes

  4. GEN 42025

    SDV Platforms

    HPC / Zonal

    The whole vehicle

    9 surfaces4 boxes

Twenty-four years, one direction. The surfaces driven climbed from one to nine while the compute stayed a handful of boxes: the software moved from the edge of the cabin to the center of the vehicle.
06Platforms · Parts list

Eight production platforms, four silicon generations.

Production platforms by program, silicon, market, and lifetime units
ItemProgramSiliconMarketUnits
01ADI Blackfin PoCHFBT, HFCK, Car Telematics Rev1/2Proof-of-concept platformsADI Blackfin––
02Holden iQ2010via ContinentalADI BlackfinAustralia150K
03Renault MM2012Radio Midvia ContinentalTI JacintoEurope3M
04Hyundai GenesisDIS 2.0via LG (LGE)Intel Tunnel CreekSouth Korea–
05JLR InControlTouch Provia HarmanIntel Bay TrailWorldwide2.1M
06Chery Exeed LX, Hongqi HS5/HS72019, 2020Automotive Workload ConsolidationIntel Apollolake, ACRNChina5M
07TTTech Nerve Blue2020Industrial Workload ConsolidationIntel ACRNEurope–
08JLR SDV platformCurrentSoftware-defined vehicleHPC / ZonalWorldwide–
07Navigation · Where the work happened

Five bases. Worldwide, across six key regions.

IndiaUnited KingdomIreland (Shannon)Chandler, ArizonaPortland, Oregon
  1. India2001-2013
  2. United Kingdom2013-2016
  3. Ireland (Shannon)2016-2017
  4. Chandler, Arizona2017-2025
  5. Portland, OregonQ4 2025-present
A place I workedCurrent baseMarkets reached
08Purpose · Why, how, what

Why any of this.

CUSTOMERHOWWHYORGANIZATION
Why
Every journey should be as inspiring as the destination.
How
Industrialize each new class of compute: production-ready, safe, manufacturable, generation after generation.

What · two outcomes

For the customer

Cockpit and SDV platforms that move the experience from vehicle-centric to customer-centric.

For the organization

Research turned into deployable, manufacturable reality at scale, and with it, technology leadership.

09Brand · The Industrializer

The Industrializer

Takes emerging compute to production-ready, safe, manufacturable reality at scale.

  1. GEN 1

    Consumer Multimedia

    2001-2013

    Media from every source of the era made production-real inside the car.

  2. GEN 2

    Cockpit SoCs

    2013-2025

    Separate boxes consolidated onto one safely partitioned system on chip.

  3. GEN 3

    SDV Platforms

    2025-present

    A couple of HPCs and zonal boxes running the entire vehicle.

Each generation folded the last one inward. Function counts went up, box counts went down, and the software moved from the edge of the cabin to the center of the vehicle.

10The Journey · Interactive

Drive it yourself

Four lane changes, twenty-five years.

Twenty-five years of automotive compute, driven rather than read. Each safe lane change moves the platform up a level of complexity, from one ECU per function toward the compute for the entire vehicle.

  1. 01MEDIA
  2. 02PREMIUM
  3. 03CONSOLIDATION
  4. 04SDV
Enter the journey
11Connect

Building something that has to ship?

I work on the part where promising compute becomes a platform that survives production, certification, and a decade of field life.