Adam Noonan

mad scientist working on calibrated uncertainty — models that know what they don’t know

Research

Calibration is a property of outputs: when you say 85%, does 85% happen? i work on earning that property honestly, auditing that it holds, and acting on it.

Effective sample size for conformal calibration under shared ancestry

Preprint. Split conformal’s coverage law assumes i.i.d. calibration data. Generalises it to correlated families — beam branches, resampled particles — and shows the design effect this literature reports is the wrong quantity, not just an imprecise one.

Coverage inversion: a calibration-transparent oracle for closed-market hours

Preprint, near submission. The consumer picks the coverage level; the oracle returns the band that empirically delivers it, with a receipt anyone can re-derive from public data. 12 years of weekend and overnight gaps. Shipped as soothsayer.

Selection on the score

In draft. When calibration-set membership depends on the score being calibrated, the correction weighted conformal applies to covariate shift doesn’t exist. What replaces it is a sharp interval.

Twin fidelity without memorisation

Fine-tuning a model to carry a person’s idiolect while it learns none of their private facts — measured with extraction canaries on one side and authorship embeddings on the other.

Certified abstention

Models represent internally that their evidence doesn’t cover a question, and answer anyway. Reading that signal so every answer is grounded or escalated, under an error budget you set instead of hope for.

Writings

Occasional posts on what i’m building and why.

Can Whisper transcribe “Rap God”?

A joke ASR benchmark for fast rap — and the discovery that the models had memorized the answers.

A fully local voice loop for my Mac

Ditching a dictation subscription for Hex + Yapper — voice in, voice out, nothing leaves the machine.

All posts →

About

Director of Data & AI at Digital Locke — individualized digital twins, and the calibration research that keeps one honest.

Technical founder with 2x profitable exits.

Previously solutions architecture at Celonis/Lenses.io and NetApp.

Taught AI engineering at Immune Technology Institute, Madrid.

Into meditation, centering prayer, kettlebells, ML & quant research, hiking, camping & SUP.

Walked the Camino de Santiago. Speaks English and Spanish.

Background

2026– Director of Data & AI, Digital Locke
2025–2026 Co-Founder, Hello Dojo AI — Exit
2025 CTO & Co-Founder, Samachi — Acquisition Exit
2021–2025 Senior Solutions Architect, Celonis / Lenses
2020–2021 Enterprise Solutions Engineer, NetApp

Education

UT Austin MS · ECU BBA

Side Projects

Experiments i’m building in public.

soothsayer

Calibration-transparent fair-value oracle for tokenized RWAs on Solana, with per-regime audit receipts.

scryer

Manifest-driven Rust data platform feeding soothsayer — versioned, deduped parquet for crypto, oracle, and market-data research.

jarvis

Personal AI ops center — realtime voice PWA over WebRTC, async research workers, and a synced Obsidian vault on Convex.

colony

Elixir + Kafka durable coordination runtime for self-healing infrastructure swarms, with replayable event fabric and bounded remediation.

saildock

Open-source autonomous docking stack for single-engine sailboats — simulation, control, and retrofit hardware design.

cto-in-a-box

Opinionated operating playbook and bootstrapper for new software products — phase-gated defaults for repos, CI/CD, ops, and agent harnesses.

Open Source

Codebases i maintain & put out in the open — some are libraries i needed and didn’t exist, others are forks i carry forward with enhanced/additional features.

ingot-db

Embeddable, transactional key-value storage engine in Rust — maintained fork of fjall-rs/fjall with deadlock fixes, custom sequence generators, and zstd compression.

lsm-tree

Log-structured merge-tree engine in Rust — fork of fjall-rs/lsm-tree with zstd dictionary compression, batch multi-get, zero-copy reads, and intra-L0 compaction.

solana-clmm-raydium

Pure-Rust, no-RPC swap math for Raydium’s concentrated-liquidity AMM — the missing Solana analogue of uniswap_v3_math.

solana-dlmm-meteora

Pure-Rust, no-RPC swap math for Meteora’s bin-based DLMM — closing the last math-crate gap among Solana’s major liquidity models.

restate-elixir

Elixir SDK for Restate’s durable execution runtime, targeting protocol V3.