GPU kernel modernization, deterministic simulation, and runtime-assured autonomy — purpose-built for DoD programs and national-lab research.
AVX-512 5.39× · + OMP 34× · + Two-pass 1.57× · cumulative 287×
We port legacy Fortran and C++ scientific codes to modern NVIDIA GPUs. SIMD intrinsics, CUDA, OpenMP, MPI, Kokkos. Every transformation is profiled against ground-truth baselines in Nsight Compute.
Reference: 287× cumulative speedup on a parallel molecular-dynamics kernel (Harvard HUIT cluster, 2025).
Read the deep-dive →Seed-deterministic synthetic-environment pipelines on Unreal Engine 5.7. Georeferenced terrain, multi-modal sensor output, neural sim-to-real refinement via Stable Diffusion XL and ControlNet.
Reference: defense aerial AI/ML training program — render-farm wall-clock 15 min → 5 s warm under bit-exact reproducibility.
Bayesian decision-consistency monitors for AI-driven autonomy under mission-level uncertainty. Black-box monitoring of autonomy commands against counter-factual replay environments.
Target: DAF26BZ01-NV008 (DoD Phase I SBIR, proposal in development).
Methodology published at arvaer.com/writing/ghostly.
Baseline (scalar) 1.00× █ + AVX-512 SIMD 5.39× █████ + OMP (24-thread, fused) 33.99× ██████████████████████████████████ + Binned two-pass 1.57× ██ ────────────────────────────────────────────────────────────────────────────── Cumulative 287.00× ████████████████████████████████████████████████████████████ →
Parallel molecular-dynamics kernel · ~9,800 atoms · 5.0 Å cutoff · 24-thread CPU baseline
Prior SBIR delivery on contracts of public record.
Phase II SBIR · $1.245M · POP July 2023 – January 2024
Engineering lead at Cymantix, Inc. on the OSINT data-collection substrate (AWS Glue / PySpark / OpenSearch) and NLP/LLM information-extraction pipeline supporting Space Domain Awareness for the U.S. Space Force / Space Systems Command. Directed a five-engineer team and an eight-month NC State University research collaboration on LLM techniques.
AWS Glue · PySpark · OpenSearch · NLP · LLM · OSINT
Public record: sbir.gov/awards/204607
Phase I SBIR · $15,013 · POP August – November 2025
Engineering contributor and proposal author at Cymantix, Inc. on the DiscoverOS platform — a Retrieval-Augmented Generative AI system for Cyber Threat Intelligence. Underlying RAG pipeline (HuggingFace, Llama.cpp, LangChain) assessed at TRL 6 in simulated scenarios with U.S. Space Force S6 (Information Dominance) per the public award abstract.
RAG · LangChain · Llama.cpp · CTI · TRL 6 · USSF S6
Public record: sbir.gov/awards/218443
All representations and certifications on file at SAM.gov.
Prominent Systems takes on a small number of select engagements each year — priority given to SBIR/STTR programs, prime-contractor subcontracts, and national-lab research partnerships. If you're evaluating HPC modernization, synthetic-data pipelines, or runtime-assured autonomy, reach out directly.
Direct: mikey@prominent-systems.com