SERIAL ALICEVERIFIABLE AI ENERGY
Benchmark Verify Docs Start free trial →
AI MODEL · v0.1 k-NN → v0.3 FP-TRANSFORMER

Alice Routerlearns energy physics from silicon.

An AI that trains on certified hardware measurements — not vendor specs, not simulations. It predicts power consumption, selects optimal batch sizes, recommends quantization, and schedules workloads based on real physics.

See the certified data → API reference
3
GPU architectures
50+
certified training points
4
optimisation objectives
0
vendor specs used
What Alice Router does

Six decisions.
All grounded in proof.

Every recommendation Alice Router makes is backed by a certificate you can verify. No black box — every prediction names the certs it used.

Pre-run energy prediction
Before you launch a job, Alice Router predicts the expected power draw and total Wh — from certified measurements at the closest certified configuration.
Batch size selection
Finds the batch size that minimises energy per token for your throughput target. At b=128 vs b=1, efficiency improves 30–80× — Alice Router tells you exactly where the inflection is.
Quantization recommendation
BF16, FP8, GPTQ, AWQ — each has a certified energy profile. Alice Router picks the format that hits your quality floor at minimum watts. AWQ saves 47% energy per token vs BF16 on Llama.
Schedule correction
Running a batch job at 3am vs 3pm can cut carbon intensity by 40–60% (Portugal grid). Alice Router shifts non-latency-critical workloads to green windows automatically.
Hardware routing
H100, H200, B200 — different models have different efficiency curves on different hardware. Alice Router matches workload to the GPU where it runs cheapest, not just fastest.
Carbon budget enforcement
Set a CO₂ ceiling. Alice Router refuses configurations that breach it and selects the next-best option — with a certificate proving the alternative was within budget.
Layer 01 — Certified training data

Every cert is a training point.

Alice Router doesn't train on datasheets. It trains on Serial Alice certificates — Ed25519-signed, Polygon-anchored measurements from real GPUs running real models. Each certificate contributes a CertifiedPoint: model, hardware, batch size, quantization, power (W), energy per token (µWh/tok), trust score.

Ed25519 signedPolygon anchoredtrust-weightedno simulations
certified-point · sa-d64b2e20TRAINING
model_id phi-3-mini-4k
hardware H200_SXM_141GB
batch_size 128
quantization BF16
avg_power_w 516.9
uwh_per_token 11.84
trust_score 0.80 hardware_attested
tee_attested true · TDX · MRTD ✓
Layer 02 — Trust-weighted k-NN

Closer certs, higher trust, more weight.

When you request a prediction, Alice Router finds the k nearest certified configurations by model similarity, hardware match, and batch distance. Each point is weighted by model closeness × certificate trust score. Higher trust (hardware_attested = 0.8) contributes more than self-reported (0.3). The prediction is a trust-weighted average — never a guess.

k-NN on cert spacemodel-similarity distancetrust × evidence
router · weight computation
request: phi-3-mini · H200 · b=128 · optimize_for=energy
cert sa-d64b2e20 w=0.80 exact match · trust 0.80
cert sa-1751b088 w=0.52 b=64 · trust 0.80
cert sa-1a051cf4 w=0.34 b=16 · trust 0.80
cert results_h100 w=0.21 H100 · trust 0.55
predicted_power_w 516.9 W
confidence 0.86 (exact cert hit)
Layer 03 — Objective optimisation

Four modes. One honest answer.

Set your objective: energy, carbon, latency, or cost. Alice Router evaluates every feasible (batch, quant, hardware) configuration against the objective and picks the winner — with progressive constraint relaxation if the strict set is empty. The decision names its basis certificates, so you can check the maths.

energycarbonlatencycostprogressive relaxation
router · decision · optimize_for=carbon
evaluating 12 (batch, quant, hw) configs…
winner b=64 · AWQ · H200
power 389.2 W (predicted)
carbon 47.5 gCO₂/h
saving −41% vs BF16 b=128
basis sa-1751b088, results_llama_quant/awq
confidence 0.68
Layer 04 — The flywheel

More runs. Better model.
Better decisions.

Every certified run adds a point to the training set. Alice Router v0.1 uses k-NN on ~50 certified points. v0.3 (coming with the next H200 rental) retrains a FP-Transformer ensemble on 200+ points including H200 real data — cutting extrapolation error from 27.6% to <10%. The model gets sharper the more you use the platform.

v0.1 k-NN · livev0.2 FP-Transformer · H100/B200v0.3 H200 data · pending
ALICE ROUTER WORKLOAD RUNS SA CERTS ROUTER TRAINS BETTER PICKS
Version roadmap

Getting smarter
with every certified run.

v0.1 · live
Trust-weighted k-NN
~50 certified points. H100 + B200 benchmark dataset. H200 TDX-certified battery. Exact model lookup + progressive relaxation. Pure stdlib, no ML deps.
v0.2 · trained
FP-Transformer baseline
Trained on H100 + B200 certified points. 27.6% extrapolation error — good on seen hardware, gaps on H200 (no real certs at train time).
v0.3 · pending H200 rental
FP-Transformer + H200 real data
Retrain with phi3/qwen/mistral/MoE/reasoning H200 certs. Target <10% extrapolation error. 8-ensemble FP-Transformer, hardware-aware feature set.
v0.4 · future
Online learning
Router updates weights continuously as new certs arrive. Tenant-specific fine-tune for your workload profile. Real-time carbon schedule optimisation.
API endpoints

One call.
A certified recommendation.

POST
/v2/router/recommend
Core router. Pass model_id, constraints, optimize_for. Returns recommended (batch, quant, hw) + predicted power + basis certificate IDs + confidence.
POST
/v2/router/plugin/vllm
Returns a copy-paste-ready vllm serve command with certified power evidence. Phi-3-mini on H200: direct TDX-certified overlay (trust=0.8).
POST
/v2/settlement/evaluate
Post-run settlement oracle. Evaluates a batch of certs against SLA conditions (energy, carbon, power, trust). Returns PASS/FAIL/PARTIAL + Ed25519-signed verdict for Polygon.
GET
/v2/router/dataset
Export the full certified training dataset as JSON. All certified points with cert IDs, so you can verify every data point the model trained on.

Every prediction.
Verifiable.

Alice Router names the certificates it used. Call /v2/certificates/{id}/verify on any of them. The energy physics behind the recommendation is on-chain — permanently.

See the benchmark data → Verify a certificate Get API access