<?xml version="1.0" encoding="UTF-8"?>
<feed xmlns="http://www.w3.org/2005/Atom" xml:lang="en"><id>https://ajliouat.com/feed.xml</id><title>Abdeljalil Jliouat — Applied AI &amp; Systems Engineering</title><subtitle>Code, experiments and engineering notes on production AI.</subtitle><link rel="self" href="https://ajliouat.com/feed.xml"/><link href="https://ajliouat.com/blog/"/><updated>2026-09-06T00:00:00Z</updated><author><name>Abdeljalil Jliouat</name><uri>https://ajliouat.com/about.html</uri></author>
<entry><id>https://ajliouat.com/blog/foundations-for-robotic-learning.html</id><title>Foundations for deep learning in robotic systems</title><link href="https://ajliouat.com/blog/foundations-for-robotic-learning.html"/><updated>2026-05-20T20:38:25+02:00</updated><summary>Practical notes on deep learning for robotics: simulation, reinforcement learning, vision-language models and the constraints imposed by real hardware.</summary></entry>
<entry><id>https://ajliouat.com/blog/observability-for-llm-systems.html</id><title>Observability for LLM systems beyond logs and metrics</title><link href="https://ajliouat.com/blog/observability-for-llm-systems.html"/><updated>2026-05-20T20:38:25+02:00</updated><summary>How to observe LLM behavior through evaluations, evidence and feedback loops, beyond the logs and resource metrics used for conventional services.</summary></entry>
<entry><id>https://ajliouat.com/blog/production-runtime-akiod.html</id><title>Designing a runtime layer for governed, vendor‑neutral AI</title><link href="https://ajliouat.com/blog/production-runtime-akiod.html"/><updated>2026-05-20T20:38:25+02:00</updated><summary>Design notes on AI runtime layers: governing tool execution, preserving evidence and managing vendor choice and cost in production systems.</summary></entry>
<entry><id>https://ajliouat.com/blog/quantum-ai-foundations.html</id><title>Sketches toward quantum AI</title><link href="https://ajliouat.com/blog/quantum-ai-foundations.html"/><updated>2026-05-20T20:38:25+02:00</updated><summary>A practical introduction to qubit states, variational quantum circuits and hybrid training loops that combine quantum and classical computation.</summary></entry>
<entry><id>https://ajliouat.com/blog/rag-pipelines-practical-notes.html</id><title>Practical notes on RAG pipelines</title><link href="https://ajliouat.com/blog/rag-pipelines-practical-notes.html"/><updated>2026-05-20T20:38:25+02:00</updated><summary>Practical RAG design notes covering retrieval objectives, indexing, chunking, hybrid scoring and orchestration, with pseudocode and an architecture diagram.</summary></entry>
<entry><id>https://ajliouat.com/blog/robotics-energy-systems.html</id><title>Robotics × energy systems</title><link href="https://ajliouat.com/blog/robotics-energy-systems.html"/><updated>2026-05-20T20:38:25+02:00</updated><summary>Thinking about robot fleets as energy systems: task assignment, charging, power constraints and policies that balance cost against service quality.</summary></entry>
<entry><id>https://ajliouat.com/blog/neural-signal-decoding-transformer-pretraining.html</id><title>Neural signal decoding via transformer pre-training</title><link href="https://ajliouat.com/blog/neural-signal-decoding-transformer-pretraining.html"/><updated>2026-05-20T20:21:50+02:00</updated><summary>Building NeuroLLM: masked EEG pre-training, motor-imagery fine-tuning, frequency-band attention and comparisons with classical BCI baselines.</summary></entry>
<entry><id>https://ajliouat.com/blog/quantum-classical-hybrid-energy-grid-optimization.html</id><title>Quantum-classical hybrid optimisation for energy grids</title><link href="https://ajliouat.com/blog/quantum-classical-hybrid-energy-grid-optimization.html"/><updated>2026-02-22T18:52:33+01:00</updated><summary>Building QuantumGrid: unit commitment as QUBO and Ising models, QAOA and VQE experiments, classical baselines, penalty tuning and scaling limits.</summary></entry>
<entry><id>https://ajliouat.com/blog/hierarchical-vlm-rl-manipulation.html</id><title>Building a hierarchical VLM+RL manipulation system</title><link href="https://ajliouat.com/blog/hierarchical-vlm-rl-manipulation.html"/><updated>2026-02-22T14:58:19+01:00</updated><summary>Inside RoboLLM: a vision-language planner, object grounding and learned motor policies for hierarchical manipulation tasks in MuJoCo simulation.</summary></entry>
<entry><id>https://ajliouat.com/blog/writing-cuda-kernels-for-transformer-inference.html</id><title>Writing CUDA kernels for transformer inference on T4</title><link href="https://ajliouat.com/blog/writing-cuda-kernels-for-transformer-inference.html"/><updated>2026-02-22T14:58:19+01:00</updated><summary>Implementing FlashAttention, fused GeLU, RoPE and paged KV-cache in CUDA and Triton, and interpreting NVIDIA T4 results with the roofline model.</summary></entry>
</feed>
