01 — The premise
Most consequential work leaves no trace. A shift is walked, a market is read, a risk is carried — and the record afterwards is a guess with a confident tone. Luna Labs builds the instruments that replace the guess with a measurement.
The moon is the fixed, silent reference by which one navigates. Every system here has one: a dock, a baseline, a known origin — the thing that makes everything downstream of it checkable.
02 — Disciplines
Four practices that keep
answering each other.
Sensor fusion borrows its uncertainty machinery from actuarial work. Trading risk borrows its regime detection from signal processing. The overlap is the point.
-
Machine learning
Sensor fusion · multimodal · retrieval
Aligned modality encoders around small language spines. Self-supervised pretraining where labels are scarce — which is nearly always. Calibrated confidence, and abstention as a first-class output.
-
Quantitative finance
Execution · risk · backtesting
Trading systems where the same strategy object runs live and in the backtester, so there is no second implementation to drift. Conviction-scaled sizing, drawdown circuit breakers, crash-safe flatten.
-
Actuarial studies
Loss models · reserving · health economics
Stochastic modelling applied to real cost bases — specialist economics, staffing demand, procedure-level margin. Distributions rather than point estimates, and sensitivity shown rather than asserted.
-
Data infrastructure
Pipelines · scale-to-zero · tooling
Ingestion that survives the source changing shape. Architectures priced for idle, not for the demo. Developer tooling that removes a class of manual work rather than accelerating it.
03 — Selected work
Systems, not demonstrations.
- 01 Ichne Every shift, proven. Sensor fusion · Multimodal ML 2026
- 02 Chronos A native quant terminal, its strategies, and the returns to show for them. Trading systems · Native + local AI 2026
- 03 Aphelion The farthest point, still in orbit. Applied LLM · Document systems 2026
- 04 Health Index Discover, enrich, and model a national health-services market. Health economics · Geospatial ML 2025
- 05 Specialist Projections How many specialists a place actually needs. Demand modelling · Health workforce 2025
- 06 Radiology Workforce Model National imaging volume, allocated to clinics, resolved to FTE. Workforce modelling · Health economics 2026
0
Systems in the corpus
0
Disciplines held at once
0
Languages in production
∞
Open arcs — the gap is the promise
04 — Method
The order matters more
than any single step.
-
01
Find the dock
Identify the fixed reference that makes everything downstream checkable.
-
02
Measure first
Audit the data as it is, not as the specification claims. This step routinely rewrites the architecture.
-
03
Compute what can be computed
Deterministic machinery for geometry and arithmetic. Learned components only where they are required.
-
04
Render the uncertainty
Every estimate ships with its error, and low-confidence output is allowed to abstain.
Most system failures are traceable to running step three before step two. The full method →
05 — Voice
Four principles,
printed on everything
we ship.
They are constraints, not values. Each one rules something out — which is the only test of whether a principle is real.
-
01
Evidence over adjectives
A claim without a measurement behind it is a mood. Numbers carry provenance or they do not ship.
-
02
Uncertainty is rendered, never hidden
Every estimate arrives with its error. A confident wrong answer is the most expensive artifact a system can produce.
-
03
Geometry is computed, language is generated
Models write prose. They do not write coordinates. Every figure a model emits is machine-checked against its computed source before display.
-
04
State negative results plainly
What failed is the more useful half of the record. Findings that contradict the plan get the same typeface as the ones that confirm it.
06 — Contact
If the measurement matters
more than the adjective —
Available for machine learning, quantitative finance and actuarial systems work. Research collaborations and open-source contributions welcome.