Senithu Dampegama — Founder & Engineer | AETRACH Industries SKIP TO CONTENT
AETRACH INDUSTRIES

AETRACH / PROFILE 001 / UNITED KINGDOM

SENITHU
DAMPEGAMA

FOUNDER & ENGINEER AETRACH INDUSTRIES LTD

Building intelligent systems that operate in the physical world.

COMPUTER VISION / AUTONOMOUS SYSTEMS / EDGE AI / ROBOTICS

EXPLORE AETRACH INDUSTRIES →
001 / PROFILE TECHNICAL DOSSIER

INTELLIGENCE
INTO MACHINES.

I work across computer vision, edge AI, autonomous navigation and embedded systems, with a focus on moving machine intelligence from models into reliable physical-world systems.

The work spans the complete engineering path: dataset development, model training and evaluation, optimisation for embedded hardware, sensor integration, navigation, simulation and real-time deployment.

ROLEFounder & Engineer
COMPANYAETRACH Industries Ltd
PROFESSIONAL FIELDDefence & Aerospace Engineering
CURRENT ENGINEERING ROLEComputer Vision Engineer
EDUCATIONBEng Robotics & Artificial Intelligence
INSTITUTIONUniversity of Hertfordshire
BASEDUnited Kingdom
002 / SYSTEM ARCHITECTURE THE SIGNAL PATH

FROM SENSING
TO DEPLOYMENT.

The systems I work on connect sensing, perception, estimation, decision-making, control and embedded deployment.

Physical systems begin with sensing. Camera, inertial and environmental data must be acquired with known timing, quality and operating constraints.

CAMERAS / IMU / BAROMETER / SENSOR INTERFACES / DATA ACQUISITION

RELATED WORK: 001 / 002 ↓

Perception converts raw sensor data into information about objects, motion and the surrounding environment.

COMPUTER VISION / OBJECT DETECTION / VISUAL ODOMETRY / MACHINE LEARNING / TRACKING

RELATED WORK: 001 ↓

Estimation combines incomplete and noisy measurements into a usable model of the system and its environment.

SENSOR FUSION / STATE ESTIMATION / FILTERING / DRIFT ANALYSIS / ENVIRONMENTAL UNDERSTANDING

RELATED WORK: 002 ↓

Decision layers translate estimated state and system objectives into a selected course of action.

AUTONOMOUS LOGIC / NAVIGATION / PLANNING / SYSTEM BEHAVIOUR / FAILURE RESPONSE

RELATED WORK: 002 ↓

Control turns selected behaviour into timed physical outputs while maintaining stability and respecting system constraints.

ROBOTICS / PX4 / REAL-TIME CONTROL / SYSTEM INTEGRATION / PHYSICAL PLATFORMS

RELATED WORK: 002 ↓

Deployment converts development-stage intelligence into an optimised system capable of operating on constrained physical hardware.

NVIDIA JETSON / TENSORRT / ONNX / EMBEDDED LINUX / REAL-TIME INFERENCE

RELATED WORK: 001 / 003 ↓
003 / ENGINEERING WORK SELECTED ENGINEERING EXPERIENCE

ENGINEERING
IN PRACTICE.

Selected areas of engineering experience across perception, autonomy and embedded deployment, presented at capability level.

001

REAL-TIME VISUAL PERCEPTION

COMPUTER VISION / MACHINE LEARNING / DATASET ENGINEERING / EDGE DEPLOYMENT

Development of machine-learning perception systems from dataset architecture and evaluation through optimisation and embedded deployment.

PATH: DATASET PREPARATION → TRAINING → EVALUATION / HARD-NEGATIVE ANALYSIS → ONNX EXPORT → TENSORRT OPTIMISATION → EMBEDDED INFERENCE

Constraints: fixed latency budgets, embedded memory and thermal ceilings, real-time camera pipelines. Validation: evaluation against realistic data with regression gates on-target, so accuracy claims survive deployment rather than benchmark conditions.

CAMERA FRAME / GEOMETRIC SCENE FIELD NORMALISED FOR THE MODEL FEATURES + DETECTION GEOMETRY T1T2 OBJECTS CONNECTED THROUGH TIME STATE / TRACKS / CLASSESUSABLE INFORMATION LEAVES THE SYSTEM
RAW INPUT ABSTRACT ENGINEERING VISUALISATION
002

GPS-DENIED AUTONOMY

VISUAL ODOMETRY / SENSOR FUSION / ROS 2 / PX4 / STATE ESTIMATION

Development of vision-led navigation and sensor-fusion systems for environments where conventional satellite positioning is unavailable or unreliable.

PATH: CAMERA + INERTIAL + BAROMETRIC SENSING → VISUAL ODOMETRY → FUSION / FILTERING → STATE ESTIMATE → NAVIGATION → CONTROL (ROS 2 / PX4)

Constraints: sensor drift and aliasing, degraded visual conditions, real-time compute limits. Validation: drift analysis, confidence-aware estimation, simulation rehearsal and controlled testing before hardware.

ORIGINABSTRACT ENGINEERING TRAJECTORIES
GNSS AVAILABLE TOGGLE LAYERS / PHASES
003

EMBEDDED INTELLIGENCE

NVIDIA JETSON / TENSORRT / ONNX / VIDEO PIPELINES / REAL-TIME SYSTEMS

Engineering and optimisation of AI pipelines for constrained compute platforms and real-time operating conditions.

PATH: TRAINING MODEL → GRAPH EXPORT (ONNX) → RUNTIME OPTIMISATION (TENSORRT) → EMBEDDED GPU EXECUTION → REAL-TIME OUTPUT

Constraints: memory ceilings, latency budgets, camera integration and deterministic timing. Validation: on-target profiling and regression checks so optimisation gains hold across releases.

DENSE DEVELOPMENT GRAPH PORTABLE STRUCTURED GRAPH LAYERS FUSED / GRAPH COMPACTED ABSTRACT EMBEDDED MODULEGPUCPUMEMORY CLEAN REAL-TIME CHANNEL ACTIVE
TRAINING MODEL ENGINEERING VISUALISATION
004 / EXPERIENCE FOUNDATION → CAPABILITY

CAPABILITY
BUILT IN LAYERS.

ROBOTICS & AI → COMPUTER VISION + EMBEDDED DEPLOYMENT → AETRACH SYSTEM CAPABILITY

AETRACH INDUSTRIES LTD CURRENT

FOUNDER & ENGINEER

Building the company's initial capability around intelligent systems, autonomy, embedded computing and applied engineering.

SENTARI LABS CURRENT

COMPUTER VISION ENGINEER — DEFENCE R&D

Developing real-time perception systems, machine-learning pipelines and edge-deployment capability for demanding operating environments.

UNIVERSITY OF HERTFORDSHIRE

BENG ROBOTICS & ARTIFICIAL INTELLIGENCE

Focused on computer vision, autonomous systems, robotics, navigation and the integration of machine intelligence with physical platforms.

005 / TECHNICAL INDEX HOVER / TAP TO TRACE

TOOLS WITHIN
THE SYSTEM.

Each tool is shown where it actually operates — in the engineering work and along the SENSE → DEPLOY path.

INTELLIGENCE

DEPLOYMENT

AUTONOMY

ENGINEERING

SELECT A TOOL

Hover, focus or tap any entry to see where it operates within the work and the SENSE → DEPLOY architecture.

006 / OPERATING PRINCIPLES 01–04

HOW THE WORK
IS APPROACHED.

01

BUILD WHAT CAN BE TESTED

An engineering claim should eventually resolve into a measurable system.

02

DESIGN THE COMPLETE SYSTEM

Perception, compute, sensing, control and operating conditions must be engineered together.

03

OPTIMISE FOR DEPLOYMENT

Laboratory performance is only one part of the problem. Real systems operate under latency, power, compute and environmental constraints.

04

LET CAPABILITY COMPOUND

Each completed system should create reusable knowledge, infrastructure and technical depth for the next.

007 / BUILDING AETRACH CAPABILITY → ORGANISATION

BUILDING AN
ENDURING CAPABILITY.

I founded AETRACH to build a long-term advanced engineering organisation around intelligent and physical systems.

The company begins with capabilities that can be demonstrated today — computer vision, autonomy, embedded AI and applied engineering — and expands through working systems, accumulated knowledge and disciplined execution.

PERCEPTION AUTONOMY EMBEDDED COMPUTE ROBOTICS ENGINEERING R&D
AETRACH emblem

The objective is not to appear larger than the organisation is. It is to continuously become more capable.

EXPLORE AETRACH INDUSTRIES

008 / CONTACT

DISCUSS A TECHNICAL PROBLEM.

For selected engineering programmes, technical collaborations and enquiries relating to AETRACH.

CONTACT AETRACH