◀ MAIN CV
👤

Profile

Nathan Timmis

Nathan Timmis

Java Platform Engineer · Android NDK · OCR & Computer Vision · Distributed Systems Architect
Full-stack Java engineer with over a decade building high-throughput distributed systems, six of them in Swiss financial services. Deep expertise in JVM performance tuning, Android NDK/JNI native integration, real-time OCR pipelines (TrOCR, PaddleOCR, GOT-OCR2), and Kafka-native event architectures processing millions of events daily. Currently shipping production JNI face-recognition engines with SIMD-optimized native code and building configuration validators for mainframe JCL pipelines at UBS. Combines JVM-ecosystem depth (Spring Boot, Kafka Streams, gRPC) with formal-methods rigour and ML/CV integration.
✉ nattimmis@gmail.com 📞 +41 78 208 84 62 🇬🇭 Urdorf ZH, Switzerland 🔗 github.com/nattimmis 🔗 linkedin.com/in/nathantimmis
☕

Java & JVM Ecosystem

Java (SE/EE)
10+ years · JDK 8-21, JNI, Concurrency
Apache Kafka
8 years · Streams, Connect, Schema Registry
Android NDK / JNI
4 years · ARM64 NEON SIMD, native C++
Spring Boot / Spring Cloud
5 years · Microservices, WebFlux, Security
Gradle / Maven
8 years · Multi-module, CI integration
JUnit / Mockito / TestContainers
6 years · TDD, Integration testing
gRPC / Protobuf
4 years · Service mesh, streaming RPCs
PostgreSQL / Redis
9 years · JDBC, connection pooling, caching
👁

OCR & Computer Vision

TrOCR / Transformer OCR
3 years · HuggingFace, fine-tuning
PaddleOCR v5
2 years · Mobile det/rec, server pipelines
GOT-OCR2
2 years · General OCR theory, VL models
InsightFace / FAISS
4 years · 512D embeddings, 152K index
TensorFlow Lite (Android)
3 years · On-device inference, quantization
OpenCV
5 years · Image preprocessing, contours
💼

Experience - Java Focus

Luxoft Switzerland AG - client UBS 22 Sep 2025 - 30 Jun 2026
PL/I Developer · z/OS market migrations
• Cut overnight batch runtime 35% across 340+ JCL jobs by profiling JVM-based scheduling agents and optimizing Java utility classes for VSAM dataset I/O
• Built OCaml-to-Java bridge validators for JCL configuration files, ensuring type-safe parameter passing between mainframe COBOL/PL-I and JVM middleware layers
• Implemented Kafka-based event sourcing for mainframe batch audit trails under FINMA / GDPR / ISO 27001 compliance, processing 50K+ events/day through Java consumers
Java 17KafkaJCLz/OSVSAMFINMAISO 27001
Raiffeisen Schweiz 01 May 2025 - 07 Jul 2025
DevOps / AI Engineer
• Designed Kafka Streams topologies for real-time ML feature extraction - Kafka data flows served the bank's internal consumers
• Built Spring Boot microservices wrapping quantized GGUF inference (Qwen2.5-3B) behind gRPC endpoints, achieving sub-200ms P99 latency under concurrent load
• Implemented OCR document processing pipeline using TrOCR + PaddleOCR for automated KYC document verification of Swiss ID cards and bank statements
• Created JUnit/TestContainers integration test suite with 94% coverage gating all ML pipeline deployments via GitLab CI
Kafka StreamsSpring BootgRPCTrOCRPaddleOCRJUnit 5TestContainersGitLab CI
Worldline 14 Mar 2022 - 31 Mar 2023
Senior Security Architect · SRE, Acquiring Online Global
• Architected Java-based certificate lifecycle management for Worldline's Acquiring Online Global platform - migration strategies designed for minimal impact on uptime on K8s
• Implemented security architecture and compliance frameworks applied in a modern, automatable form to newly created payment infrastructure
• Built Prometheus JMX exporter dashboards monitoring GC pause times, heap allocation rates, and thread pool saturation across the certificate authority stack
Java 11K8sSecurity architectureSREPrometheusJMXPaymentsHelm
🚀

Flagship Projects

i-XRAY - Real-Time Face Recognition Engine (Android NDK + Java JNI)
Enterprise-grade facial recognition system with Java JNI wrapper over C++ SIMD-optimized native library. Singleton NDK engine with searchUltraFast() achieving sub-millisecond identification across 152K-face FAISS index. ARM64 NEON vectorized dot products, memory-mapped database loading, and Unit 8200-style threat/priority scoring. TFLite mobile deployment on Samsung S24 with real-time Meta Ray-Ban camera feed processing at 100+ faces/minute.
Java JNIAndroid NDKC++ SIMDFAISSInsightFaceTFLiteARM64 NEON
IXrayEngine.java - JNI Native Bridge
package com.ixray.ndk; public class IXrayEngine { static { System.loadLibrary("ixray_ndk"); } // Native methods - C++ SIMD backend public native boolean loadDatabase(String dbPath); public native MatchResult[] searchUltraFast(float[] embedding, int topK); public native double benchmark(int iterations); // High-level identification public IdentificationResult identify(float[] embedding) { MatchResult[] matches = searchUltraFast(embedding, 5); if (matches == null || matches[0].confidence < 45.0f) return null; return new IdentificationResult(matches[0]); } }
Drone-Sense - Triple-Detection Drone Sensing (Kotlin + Android)
WiFi/BLE RemoteID detection + Acoustic FFT drone identification on ESP32 with Android AR overlay. Kotlin Android application with JNI-bridged C++ signal processing for real-time FFT analysis. Multi-sensor fusion combining RF fingerprinting, acoustic signatures, and visual detection via on-device TFLite models.
KotlinAndroidJNIC++ESP32TFLiteFFTBLE
JobScan24 - Swiss Job Intelligence (Kafka + Java Backend)
Hybrid CPU/GPU inference platform: BGE-large (1.3 GB) embeddings + Qwen2.5-3B ranking across 30+ API endpoints. Kafka-native job event pipeline with Java consumers for real-time job matching and alert distribution. Dual VPS / on-prem RTX deployment with Java-based orchestration layer.
KafkaJavaBGE-largeQwen2.5gRPCPostgreSQLRedis
SixthSense - Wearable AR System (Android + Python)
Dual-implementation wearable augmented reality system: Android app with camera-based gesture recognition and Python backend for real-time projection mapping. OCR-integrated badge/credential reading for conference and security applications.
AndroidPythonOpenCVOCRARGesture Recognition
K8s Platform Toolkit - JVM Microservices Infrastructure
Reference Terraform + Helm + ArgoCD templates for AKS/EKS dual-cloud under Swiss banking compliance. JVM-optimized container images (GraalVM native-image, JLink custom runtimes), Prometheus JMX exporters, and Grafana dashboards for GC/heap/thread monitoring across 3-environment promotion pipeline.
K8sHelmArgoCDGraalVMPrometheusJMXTerraform
📄

OCR Pipeline Architecture

Document Processing Pipeline (Production)
1. Ingest → 2. Preprocess → 3. OCR Extract → 4. NER / Validate → 5. Store
Stage 1 - Ingest: Kafka consumer (Java) receives document events from upstream services (PDF, images, scans)
Stage 2 - Preprocess: OpenCV deskew, binarization, noise removal. Java ImageIO + native JNI for high-throughput batch processing
Stage 3 - OCR Extract: Multi-model ensemble: TrOCR (printed text), PaddleOCR v5 (multilingual), GOT-OCR2 (complex layouts). Java orchestration via gRPC to Python inference servers
Stage 4 - NER / Validate: Named entity recognition extracts structured fields (name, DOB, address, ID number). Confidence scoring with human-in-the-loop for <90% confidence
Stage 5 - Store: PostgreSQL with JSONB + full-text search. Kafka event emitted for downstream consumers (compliance, audit, matching)
🎓

Education & Certifications

MSc Advanced Security & Digital ForensicsEdinburgh Napier
PG Computer ScienceStaffordshire
Cert Cyber Security & Networking (Merit)Oxford
LLB Law & International RelationsKeele
AWS Solutions Architect Associate2022
Azure Administrator (AZ-104)2021
Red Hat OpenShift Specialist2020
Certified Splunk Power User2023
GCP Associate Cloud Engineer2019
PRINCE2 Practitioner2017