home
download pdf
Jonathan Tollefson
Design Engineer
jtollefson123@gmail.com · (614) 403-4845 · uxjon.com · linkedin.com/in/jtollefson123 · Minneapolis, MN

Summary

Design engineer working the seam between interface design, front-end, and applied AI. Sole designer, builder, and shipper of an agentic audio-mastering system, and of the blind evaluation harness that gates it, built after hitting the known failure mode of these systems: the evaluator, not the model, sets the ceiling on quality. Before that, enterprise AI at a Fortune 500 cooperative, a B2B SaaS AI platform, and a bilingual banking launch that cut onboarding bounce from 40% to 25%, designing with engineers rather than handing off.

Selected Builds

Finishable — the Engineer · AI mastering agent
2026 – Present
  • Built an AI mastering system that works as an agent, not a button: it takes actions and carries state across a multi-step run, rendering several masters, running a blind, loudness-matched A/B (ITU-R BS.1770, true-peak controlled), and composing a final master from the strongest takes.
  • Built an active-learning model of my own taste, then shelved it: blind A/B results showed it overfitting a biased evaluator, the known reward-model overoptimization failure.
  • Moved the work into the judge instead: a blind, loudness-matched harness that gates every change against a baseline of doing nothing. Single-rater so far; a multi-rater blind run is the next gate.
  • Designed and shipped end to end, no handoffs: Astro, React, TypeScript, Python, FastAPI.
Interactive case study: uxjon.com/case-studies/the-console
Finishable — the Catalog · audio catalog and search
2026 – Present
  • Built a searchable catalog of roughly 20,000 audio files, read by sound, fingerprinted, deduplicated, and auto-classified through a Whisper, Claude, and librosa pipeline. Sole engineer across ingestion, classification, and interface.
SIGNAL · browser instrument
2026 – Present
  • Built a playable instrument that runs in a browser tab: Web Audio synthesis, sequencing, and a performance layer, with the sound engine and the interface designed as one system. It lives on my homepage; press power.
Play it: uxjon.com/#play

Experience

CHS Inc. · Business Analyst, AI & BI Engineering
Aug 2023 – Present
  • Designed and prototyped interfaces for enterprise AI and BI products at a Fortune 500 agricultural cooperative, embedded with engineering on consumer agri-tech tools and internal AI tooling.
  • Designed a FigJam-based documentation and alignment system adopted as the cross-team default across product, engineering, and business.
  • Built and shipped a Python tool that parses the dbt manifest and fans model lineage into Excel, Power BI, and stakeholder briefs in one pass; owned it end to end for business users, engineers, and the AI team.
Raylu, Inc. · Product Designer (Freelance)
Jan – Oct 2023
  • Designed the interface, component set, and interaction model for a B2B SaaS AI chat platform, specified so the patterns mapped 1:1 to the React implementation.
  • Ran the roadmap for a distributed team shipping prototype releases on a tight cycle.
Crediverso · UX Design Lead
Feb 2021 – Jan 2023
  • Designed the primitives, patterns, and onboarding flows for a bilingual mobile banking V1, shipped on iOS.
  • Ran usability testing with 50+ participants; the iterated primary flow cut bounce from 40% to 25%.
  • Grew the pre-launch channel from 700 to 2,100 followers on A/B-tested messaging.

Education

Harvard University
2022
A.B., Sociology
Dean's List (6×) · Rosenkrantz Discovery Grant
University of Minnesota
2016 – 2018
Product Design, Business Marketing Education

Skills

Build: TypeScript, React, Astro, Python, FastAPI · Design: Figma, Framer, Adobe Creative Suite
Models & Audio: Web Audio API, Claude, Whisper, librosa, demucs · Data: SQL, dbt, Power BI
OutsideDiary of a Soundbender, a Substack on sound and product design. Track and field record holder, University of Minnesota.
how one run moves

Five steps. The last one usually says no.

  1. 1
    Render 4 to 6 masters, 4.2 on average
  2. 2
    Blind A/B loudness-matched, labels hidden
  3. 3
    Weigh votes become win-rates
  4. 4
    Champion composed from the winning takes
  5. 5
    Add more processing? usually no
    42 of 78 decided runs kept the original 54%
229 runs · 939 renders · 42 originals kept
harness log, snapshot 2026-08-26 · single rater

the harness ledger · snapshot 2026-08-26

The taste model was overfitting my ratings, not improving the audio. I shelved it and built a blind judge.

each round: two takes, unlabeled, matched to the same LUFS, judged on one axis (focal, dynamics, space, raw vs processed). the original is always on the ballot.

the ledger 10 of 50 rounds
Ten rounds from the blind listening ledger: date, axis, the LUFS both takes were matched to, and which take won. Listening notes are quoted under their round.
date axis matched lufs winner
2026-06-10 focal -15.66 render
focal -19.56 original
dynamics -21.37 original
the big bass stabs at the end, 2:30 and following, are tamed a bit in mix 2 and 3
space -21.37 original
2026-06-11 space -17.13 original
i like mix 2 for the first 35 seconds, but mix 1 wins for the drop, i.e. everything after
focal -16.53 render
space -16.16 original
i like 1 and 3 the best until the beat drops at 1:07 - then mix 2 is the best
focal -18.92 original
raw_vs_processed -15.58 render
2026-08-16 focal -16.30 original

the first nine rounds and the latest. behind them: 229 runs, 939 renders, 4.2 per run (6 at most).

58
blind loudness-matched votes
26
kept the original
single
rater

how often the original won

blind votes 26/58
decided runs 42/78

dashed line: 50%

of 78 runs that reached a champion, 42 champions were the original. doing nothing won 54% of decided runs.

single rater so far, 50 rounds across 6 sessions, 2026-06-10 to 2026-08-16. a multi-rater blind run is the next gate. no catch trials yet.

the mechanism, on sample data

One manifest in, three documents out

sample models; the real one runs on the company manifest

manifest.json jaffle-shop sample → parse → 6 models · 6 edges · 3 layers
depth 0
depth 1
depth 2

hover a model to trace its lineage ←

fans out to Excel Power BI briefs
web audio, in this page

The instrument claim, pressable.

sound on press

Synthesis, envelope, and filter run right here, in the résumé. The full instrument lives at uxjon.com/#play.

jtollefson123@gmail.com minneapolis, mn