Engineer · 14 years · TypeScript, Python, Go, AI infrastructure
fortissimo.io · me@brdjx.com · linkedin.com/in/brdjx
Relocating to the San Francisco Bay Area
Companies bring me in before the problem is defined. I run discovery with whoever owns the business outcome, build the system, and hand it off with the docs and tests their own engineers work from. More than fifty engagements across twelve industries, every one of them with me as the sole technical owner. Four of the products were later acquired.
The work runs through Fortissimo, my own studio. Boutique by design rather than an agency: there is no bench and no account layer, so the engineer you talk to is the engineer who builds it. That is the same person on all fourteen years of the work below.
Why this profile is empty. Almost everything I have built belongs to a client and lives in a private repository. That is the ordinary shape of a freelance career, and I would rather say so than pad this page. What I can point you at is below, and most of it you can inspect yourself.
SoundMint · Creator-to-fan music platform I founded and built. The ownership protocol is deployed to Base mainnet, settles in USDC, and control is held by a multi-signature safe, so no single party, including me, can change the terms artists agreed to after they agreed to them. On-chain and verifiable. Open beta with a 120-artist cohort.
UltimateFightingStats · A win-probability model I trained, calibrated, and serve myself. 81 features across 8,464 historical fights, temporal-split validation and Platt calibration, landing at 65.8% accuracy and 0.716 AUC on a fully held-out window. It runs as ONNX inside the API, so there is no Python in the request path. Promotion is gated on calibration: a degraded model still trains and uploads, but it can never reach a user.
Lineups.com · Catena Media, 2022
The Nation Network / DailyFaceoff · Playmaker, 2019
Targetable · Chowly, 2023
Project DADO · Stanley X, 2022
Built under contract. I was the builder rather than an equity holder, so the exits validate the bets, not my bank account.
Fundamental Research Labs · ML deployment and evaluation platform · contract, 2025
Built the model deployment and evaluation platform for the a16z-backed lab behind Shortcut.ai, from first commit to a working MVP the research team was ramping onto, in about two months.
Founded its evaluation service. The interesting part was making twelve or more model families usable behind one interface, which meant auto-detecting context windows and tool-call formats rather than hardcoding them per model. Dataset runs went through a scheduler I wrote with bounded concurrency and per-batch checkpoint and resume, built on the assumption that the process would be killed mid-run, because it was.
Angeles One · the whole operating platform, solo, in four months · contract, 2025 to 2026
Hired to add two integrations to a music school's existing no-code CRM. I spent the first week with the founder and the front desk instead of building, and concluded the tooling was the problem rather than the integrations. So I proposed replacing the stack.
Delivered end to end and solo: 34 Lambdas, roughly 150 API routes, 25 Postgres tables across 99 migrations, and a React front end of 155 components across 30 screens, TypeScript throughout. Every call, text, voicemail, and email is captured exactly once and in order, matched to the right family, and summarized, so the front desk works from one clean timeline instead of hunting duplicates across four systems.
More than 1,500 weekly students across four locations run on it, at roughly ten dollars a month of infrastructure. I built it so their own junior developer could maintain it after I left, which is how it still runs.
GoodPix · modernized an inherited platform with no cutover · contract, 2020 to present
Frontend contractor to fractional CTO over six years. I inherited a first-generation product whose backend was 27,000 lines of monolithic classes with no model layer, and whose front end took three minutes to hot reload.
A rewrite was the obvious move and I decided against it, because it would have frozen product work for months and staked the business on a single cutover. Instead I stood a v2 alongside v1 inside the same application so both generations served live traffic, gave v2 the model layer and service layer and test suites it had never had, and wrote migration routes that moved live users across incrementally. There was never a cutover weekend.
Top committer on six of the nine repositories, and on both the backend and the front end of the two largest.
Lineups.com · founding frontend platform, still in production nine years later · contract, 2017
Established the entire frontend platform in four months and built it search-visibility-first, choosing server-side rendering when that was still difficult, because crawlable content was the growth channel.
That architecture became the company's organic growth engine across five sports verticals and absorbed years of expansion without a rewrite. It was still in production through the 2022 acquisition.
TypeScript, Python, and Go in production, with React on the front end and native mobile in Swift and Kotlin when a project needs it.
Most of the last year has been AI infrastructure: deployment and evaluation platforms, agent tooling, and the unglamorous half of making LLM systems reliable enough that people can actually depend on them. When an agent system I operated fabricated a URL, the fix was not a better prompt. I changed the service so it raises an error rather than emitting anything it cannot trace back to evidence. I build with agentic coding tools daily and have opinions about where they break.
Outside work I produce music, and I co-founded and coach a FIRST Robotics team at a small rural school in Michigan's Upper Peninsula.