I design and deploy high-performance computer vision systems for real-world industrial environments.
- ⚡ Real-time pipelines (30+ FPS) under constrained hardware
- 🧠 GPU acceleration (CUDA / TensorRT / zero-copy)
- 🏭 Industrial vision systems (inspection, automation, edge deployment)
- 🔧 Robust perception under non-ideal conditions (lighting, noise, reflections)
- 🚀 Optimized production pipelines from ~14 → 30 coins/sec
- ⚙️ Designed lock-free, zero-copy architectures for deterministic performance
- 🧠 Built multi-stage CV pipelines: segmentation → orientation → classification → defect detection
- 📡 Deployed systems on Jetson (edge AI) with real-time streaming and multi-client support
I specialize in systems where hardware meets AI, focusing on:
- Deterministic real-time processing
- GPU memory optimization (cudaHostRegister, zero-copy)
- Scalable vision architectures (multi-process, shared memory)
- Industrial communication (CAN / J1939 / Modbus)
Real-time inspection system built for high-throughput production lines.
- 🧩 C++ camera producer + Python consumers (shared memory)
- ⚡ Zero-copy GPU pipeline
- 🔄 Lock-free double buffering (0 ms overhead)
- 🧠 CV pipeline: segmentation → alignment → classification → defect detection
- 📡 WebSocket streaming to multiple clients
fromtypingimportDict, ListclassEngineer:
def__init__(self):
self.name="Gerald Cainicela"self.role="Embedded & CV Engineer"self.location="Lima, Peru 🇵🇪"self.focus= [
"Real-Time Computer Vision",
"Edge AI & Inference Optimization",
"Industrial Embedded Systems",
"Protocol Engineering (CAN/J1939)"
]
defget_stack(self) ->Dict[str, List[str]]:
return {
"vision": ["CUDA", "TensorRT", "YOLO"],
"embedded": ["ESP32", "FreeRTOS", "STM32"],
"edge": ["Jetson", "Docker", "OpenVINO"],
"plc": ["CODESYS", "IEC 61131-3", "Modbus"],
}
asyncdefdeploy(self, target: str) ->str:
returnf"Optimized & deployed to {target}" | #include<memory>
#include<vector>
#include<string>
#include<map>classEngineer {
const std::string name_{"Gerald Cainicela"};
const std::vector<std::string> focus_{
"Real-Time CV", "Edge AI",
"Industrial IoT", "Protocol Eng."
};
public:autogetStack() constnoexcept {
return std::map<std::string,
std::vector<std::string>>{
{"vision", {"C++20", "CUDA", "TensorRT"}},
{"embedded", {"ESP-IDF", "FreeRTOS", "ARM"}},
{"edge", {"Jetson", "Docker", "ONNX"}},
{"comms", {"CAN", "J1939", "Modbus"}}
};
}
template<typename T>
[[nodiscard]]autooptimize(T&& pipeline) {
return std::make_shared<std::decay_t<T>>(
std::forward<T>(pipeline));
}
}; |
Specialized in real-time computer vision and industrial embedded systems. I build high-performance pipelines for edge inference, industrial bus communication, and robust production deployments.
🛠️ Tech Stack
Computer Vision & AIC++14/17/20CUDAcuDNNTensorRTONNX RuntimeOpenCVOpenCV C++YOLOYOLOv8DeepStreamQt6OpenGLlibtorchEigenManim
Edge AI & DeploymentJetson OrinJetson NanoJetPackL4TDockerNVIDIA RuntimeTensorRTDeepStreamONNXOpenVINO
Embedded & IoTCC++ESP32ESP-IDFFreeRTOSMicroPythonArduinoSTM32ARM CortexRaspberry PiLoRa (SX127x/SX126x)MQTTModbus RTUCANJ1939RS485RS232OTABootloader
Industrial Automation & PLCCODESYSIEC 61131-3Ladder LogicStructured TextFunction Block DiagramModbus TCP/RTUOPC UASCADA
Protocols & Hardware InterfacesI2CSPIUARTPWMADCDMABLEWiFiESP-NOWNVSWebSocketHTTPTCP/UDPEthernetUSBProfibusProfinet
ML/DL FrameworksPyTorchlibtorchTensorFlowTensorFlow LiteONNXKerasOpenCV DNNscikit-learnNumPyPandasPlotlyMatplotlib
DevOps, Tools & LanguagesDockerdocker-composeKubernetesCI/CDGitHub ActionsGitLab CILinuxUbuntuGitCMakeMakefileBashGStreamerFFmpegGDBValgrindPerfNVIDIA NsightMATLABLaTeX
DatabasesPostgreSQLMySQLSQLiteSQL Server
CAD & DesignAutoCADSolidWorksEasyEDA
Lima, Peru · 2026



