Focused areas of work: from model compression to hardware dataflow, from embedded AI to reproducible pipelines. Case studies and experiments published as structured narratives.
This page will expand with curated research threads and case studies.
End-to-end workflow summaries from model training through compression to FPGA bitstream generation.
Real-world deployment examples on Zynq, Ultra96, and 7-series devices with synthesis results and performance metrics.
Quantization, pruning, distillation benchmarks across different model architectures and target platforms.
Split learning and edge intelligence projects exploring distributed inference and on-device optimization.
Profiling results comparing resource utilization, latency, and power across FPGA families and configurations.
Reproducible examples and structured exercises for teaching ML-to-hardware concepts.
A curated archive of past research papers, tools, prototypes, and experiments will be added later, aligned with KaleidoForge's mission and ML-to-hardware focus.