30% / 40%
Higher accuracy, lower compute
Multi-task perceptionDeveloped multi-task architectures for 2D/3D pose estimation and object detection, improving accuracy by 30% while reducing computational cost by 40%.
AI ENGINEERProduction AI · Ankara
AI Engineer turning
advanced models into
reliable production systems.
8+ years of experience across computer vision, multimodal AI, agentic workflows, real-time edge inference, and large-scale image pipelines.
8+ yearsProfessional AI experience
30 FPSReal-time edge inference
1B+Images processed in ML pipelines
COMPUTER VISIONMULTIMODAL AIAGENTIC SYSTEMSMLOPSEDGE INFERENCESEMANTIC SEARCH
01 — SELECTED IMPACT
Selected engineering outcomes. Accuracy, efficiency, and scale.
Representative results from work on multi-task learning, real-time edge inference, and large-scale vision datasets.
View selected work30 FPS
Real-time edge performance
Edge inferenceOptimized machine-learning inference on edge devices to 30 FPS, achieving a threefold performance improvement.
1B+
Large-scale vision data
Large-scale ML data pipelinesDesigned and operated machine-learning data pipelines handling more than one billion image records.
02 — SELECTED PROJECTS
Selected engineering projects.
Technical projects exploring efficient language-model inference and the delivery of complex native applications through modern web runtimes.
IN-BROWSER AI2026
1bitllm
A browser-native playground for quantized Bonsai language models. It runs ONNX inference with WebGPU, streams generated tokens from a Web Worker, and caches model assets locally without a server-side inference backend.
Engineering contribution
Designed and implemented the client-side inference flow, model lifecycle, streaming chat interface, and WebGPU capability checks.
WEBASSEMBLY PORT2019
OpenTTD / Browser Port
A browser port of OpenTTD compiled with Emscripten and WebAssembly, bringing the C++ simulation engine to a full-screen web canvas without a native installation.
Engineering contribution
Ported and packaged OpenTTD for browser execution, integrating the WebAssembly runtime, asset bundle, loading states, and canvas-based application shell.
03 — EXPERTISE & TOOLKIT
Technical depth across the production ML lifecycle.
Experience across the machine-learning lifecycle, from model development and optimization to system integration and production delivery.
01
Computer Vision
Development of image-processing, object-detection, and 2D/3D pose-estimation systems for real-time and research applications.
DetectionPoseImage processing
02
Multimodal Generative AI
Design of systems that combine image and language models for context-aware generation, analysis, and product workflows.
Vision-languageGenerationFusion
03
Agentic AI
Architecture of tool-using agents that execute multi-step workflows through external services and APIs.
AgentsTool useOrchestration
04
Production ML & MLOps
Containerized services, reproducible model delivery, and engineering practices that support reliable iteration and deployment.
MLOpsDockerMicroservices
05
Edge AI
Model and inference optimization for constrained devices, real-time video pipelines, and IR-assisted perception systems.
OptimizationCUDAIR sensors
06
Semantic Search
Compact indexing and multimodal retrieval across image, audio, and natural-language queries.
RetrievalIndexingMultimodal
Languages, tools, and specialization.
PythonC++CMATLABSQL
PyTorchTensorFlowMXNetNumPyCUDA
DockerGitMLOpsMicroservices
Computer VisionMultimodal AIAgentic AIEdge AISemantic Search
Turkish · NativeEnglish · Professional
04 — BACKGROUND
Education, research, and recognition.
An academic foundation in computer engineering, supported by machine-learning research, a conference publication, and academic distinctions.
Middle East Technical University
Graduate studies in Computer Engineering
Bilkent University
B.Sc. in Computer Engineering
A.T. Albayrak, İ. Atıl, “Hyper Parameter Optimization for Deep Learning using Volunteer Computing,” 6th High Performance Computing Conference, 2020.
- Bilkent University Senior Project Data Science Award (2017–2018)
- Ranked 753rd in the quantitative track among approximately 1.8 million candidates in Türkiye’s national university entrance examination
- Enderun Science High School Salutatorian
CONTACT
Email meahmetahalbayrak@gmail.com↗Open to international AI engineering opportunities.
For roles and collaborations involving production computer vision, multimodal AI, agentic systems, and machine learning infrastructure.
Ankara · Remote · International