Çankaya, Ankara, Türkiye
I design AI agents you can actually trust — autonomous, auditable, and safe by construction. AI / agentic-systems engineer and architect specializing in LLMs, multi-agent orchestration, AI safety, and local-first / on-device AI. I'm the creator of NEXUS, a post-quantum-immune agentic operating system, and I build in the open through the DoppleGround foundation. The problem I solve: most AI agents are powerful but ungoverned — unauditable, insecure, and prone to hallucination. I build the opposite. In NEXUS, every agent action is proposal-bound, execution-gated, and provenance-tracked on an immutable audit ledger, secured with post-quantum cryptography (ML-KEM-1024, ML-DSA-65) and protected by token-budget and hallucination guards — all running locally on low-VRAM hardware. What I work on across the AI stack: • Agentic AI — multi-agent swarms, agent-to-agent (A2A) governance, autonomous workflows, speculative routing, DAG intent planning • AI safety & security — adversarial robustness, multi-turn alignment-failure mitigation, hallucination detection, post-quantum cryptography for AI • Local & on-device AI — Ollama, GGUF quantization, WebGPU compute kernels, edge runtimes (LiteRT / TFLite) • Generative & visual AI — agent-driven image pipelines (NEXUS Visual Weaver) • Research & ML engineering — 140+ open-source repositories spanning agents, kernels, and applied LLM systems My edge is a physicist's first-principles mindset. M.Sc. Applied Physics (TU Dresden) and B.Sc. Engineering Physics, with semiconductor and nanotechnology R&D. I even apply Landau–Ginzburg phase-transition theory to detect LLM hallucination — bringing experimental rigor and Six Sigma discipline to AI that must hold up under real-world load. Core skills: Agentic AI · LLMs · Generative AI · Multi-Agent Systems · A2A Governance · AI Safety & Alignment · Adversarial Robustness · Post-Quantum Cryptography · Local-First / On-Device AI · Ollama · GGUF · WebGPU Kernels · Edge ML · AI Systems Architecture · Python 📬 Open to roles and collaboration in agentic AI, AI safety, and local-first AI infrastructure. Message me here, or explore my work at github.com/specimba and huggingface.co/specimba.
Founder of the DoppleGround foundation and architect of NEXUS — a local-first, post-quantum-immune agentic operating system for governed, auditable AI. What I build: • Agentic AI infrastructure — multi-agent swarms with agent-to-agent (A2A) governance, where every action is proposal-bound, execution-gated, and provenance-tracked on an immutable audit ledger. • AI safety & security — adversarial robustness, multi-turn alignment-failure mitigation, hallucination detection, and post-quantum cryptography for AI (ML-KEM-1024, ML-DSA-65). • Local & on-device AI — Ollama, GGUF quantization, WebGPU compute kernels, and edge runtimes (LiteRT / TFLite) tuned for low-VRAM, on-device performance. • Generative & visual AI — agent-driven image and visual generation pipelines (NEXUS Visual Weaver). • Open source — 140+ public repositories spanning agents, kernels, datasets, and applied LLM systems. I bring a physicist's first-principles approach to AI engineering — including applying Landau–Ginzburg phase-transition theory to model and detect LLM hallucination. Keywords: Agentic AI, LLMs, Generative AI, Multi-Agent Systems, A2A Governance, AI Safety, Post-Quantum Cryptography, Local-First / On-Device AI, AI Systems Architecture, Python.
Independent operation of machines for producing organic semiconductor devices, including front-end (preparation) tasks. Back-end (post-processing) detailed component characterization and documentation. Photophysical characterization of emitter materials for suitable device structures. General tasks of laboratory maintenance and optimization of throughput/quality of component production
Contributed significantly to the ongoing reactive magnetron thin film coating (PVD) project, achieving notable advancements in optimizing complex compound layers. Utilized expertise in nanoparticle source sputtering techniques to successfully coat nanomaterials with compound layers, enhancing project outcomes. Collaborated effectively with multidisciplinary teams, ensuring seamless integration of processes and technologies. Recognized for contributions and given opportunities for professional development and growth.
Developed a novel reactive magnetron sputtering (PVD) part, enhancing the nanocomposite layer production process and optimizing overall efficiency. Led the optimization efforts for TiO2 thin film coating, improving coating quality and performance through meticulous process control and monitoring. Demonstrated strong attention to detail and adherence to quality standards, successfully attaining sustainable TiO2 layers through a precise process-controlled (PID) system reactive coating progress. Authored a comprehensive master's thesis, effectively documenting academic and engineering achievements. Benefitted from a collaborative and supportive work environment, receiving guidance and mentorship from experienced professionals. Demonstrated proficiency in scientific and optical acquisition tools (OED, collimator design, fiber-optic connection arrangement for high vacuum), ensuring accurate device modification, installation, and measurement techniques (Ellipsometry, Profilometry, UV-Vis RT). Project leader references, and the thesis will be provided upon request.
Leveraged scientific programming capabilities and data visualization skills to meet industry needs effectively. Explored new materials for liquid metal battery technology and simulation, contributing to innovative solutions. Developed mathematical models and simulations to analyze and optimize elemental capabilities, enhancing battery performance. Thrived in a stimulating and collaborative work environment, where recognition of contributions and opportunities for professional growth were present.