Skip to content

Ibteshamul
Haque

CUDA kernels and LLM inference systems — and the measurement discipline to tell when something actually got faster.

Fig. 01 min–max

Horizontal range-bar chart of custom CUDA kernel speedup against a baseline. 3 comparisons; each bar spans the measured minimum to maximum multiplier across independent sessions, so bar width is run-to-run spread, not error. A reference rule marks 1×, meaning no speedup. Each comparison and its exact range follow.

  • vs framework baseline
  • vs vendor library
  • 1× baseline
Blocks 1/2/4 vs matched PyTorch · RTX 3050 3.24–6.55×
Block 3 naive → FP16 half2 gate packing 7.55–9.50×
FP16 vs cuDNN same operator 1.30–1.47×
0 2 4 6 8 10
1× baseline

speedup — × versus baseline · linear scale · higher is faster

  • Ranges span independent sessions. The spread is run-to-run variation, not error bars.
Custom kernels vs baseline · operator-for-operator · RTX 3050 COLIDE

COLIDE case study

Table 1 last 365 days

GitHub contribution calendar for titoatwork. 837 contributions in the 365 days from 7 Sept 2025 to 6 Sept 2026, on 72 days. The busiest day was 26 Jul 2026 with 98. Counts come from the public profile, so private contributions appear only if that account exposes them. This is a record of activity on one host, not a measure of work done.

Contributions
837
last 365 days
Active days
72 / 365
days with at least one
Busiest day
98
26 Jul 2026

Counts what the public profile exposes. Private contributions appear only if that account chooses to show them.

Public GitHub activity for @titoatwork · 7 Sept 2025 – 6 Sept 2026 · an activity record, not a measurement of work as of 6 Sept 2026
Table 2 public activity

No public pushes in the window GitHub reports. The feed is capped at 90 days and 300 events, so quiet periods and private work both read as nothing here.

Public push events · repository and branch only — GitHub no longer exposes commit messages in this feed as of 6 Sept 2026

Finishing COLIDE — custom CUDA kernels for an IoT intrusion-detection model, now in manuscript prep — and building Slipstream, an LLM serving engine organised around a memory-aware scheduler.

  • 01 Custom CUDA kernels for small, control-flow-heavy models
  • 02 LLM inference systems — KV-cache scheduling, quantized decode
  • 03 On-device and air-gapped deployment
  • 04 Benchmark methodology: ranges, not hero numbers

Selected work

4 projects 2 with published figures

All projects

Measurement discipline

Re-running the same benchmark suite twice in one sitting moved framework latencies by 14–17%, and moved individual kernel configurations by up to 27% — while each session's own coefficient of variation looked perfectly tight. Within-session variance understates the real uncertainty on a WSL2 development box. Every latency figure here is therefore a range across independent sessions, and single-run tables are treated as unpublishable.

Fig. 02 mean ± published spread

Point estimate with an error bar. BoT-IoT macro-F1 is 0.9780, plus or minus 0.0033, n = 5 (seeds 42–46). The interval therefore runs from 0.9747 to 0.9813. The horizontal axis is truncated: it runs from 0.96 to 0.99, not from zero. Comparators drawn for reference: protocol LightGBM · 0.9818 at 0.9818. No distribution curve is drawn, because the underlying samples are not published.

BoT-IoT macro-F1 0.9780 ± 0.0033

n = 5 (seeds 42–46)

0.96 0.97 0.98 0.99
protocol LightGBM · 0.9818 0.9818

BoT-IoT macro-F1 · linear scale · truncated axis — window 0.96 – 0.99, not zero-based

0 1

full 0–1 scale · the box marks the window plotted above

  • Interval spans 0.9747 to 0.9813: the published ± figure, plotted as published. n = 5 (seeds 42–46). The source does not state whether it is a standard deviation, a standard error or a confidence interval, so it is not labelled as one here.
  • No distribution is drawn. There is a mean and a spread, not a set of samples — a curve over these two numbers would be a shape nobody measured.
Sealed multi-seed test protocol · BoT-IoT · n=5 COLIDE

This section is load-bearing, not modesty.

  • There is no full-pipeline speedup number.

    "Custom CUDA is N× faster than TensorRT" would be an easy sentence to write and it would be false.

  • A negative result stands unretracted.

    Rather than quietly dropping the comparison, the active claim set treats server Block 3 as unresolved until a fresh run exists.

  • An earlier ToN-IoT result was withdrawn.

    The corrected leakage-safe rerun is weaker and honest — CNN 0.8075 against a same-split random forest at 0.9626 — and the old figures are tombstoned in the repository rather than deleted.

Quoted verbatim COLIDE — what is not claimed

Recent notes

Full index

Questions and corrections

A short technical note is welcome — a question about a kernel, a measurement, or something here that looks wrong. If a figure on this site does not reconcile with its source, that is a bug and I would like to know.