Hi! I'm Jiri (Jirka) Borovec — an open-source ML/CV systems maintainer and consultant with a Ph.D. in Medical Imaging.

Over the years I've maintained projects like PyTorch Lightning, TorchMetrics, Supervision, and many others. I believe great software should be open, well-tested, and accessible to everyone.

Your sponsorship helps me dedicate more time to:

• 🔧 Maintaining and improving open-source libraries
• 📝 Writing documentation, tutorials, and blog posts
• 🐛 Triaging issues and reviewing community PRs
• 🚀 Building new tools for the ML/AI ecosystem

Ways to sponsor:

Even a small contribution keeps the lights on and the CI green 💚

Who sponsors, what was delivered, and what the money funded is recorded in SPONSORS.md.

I help teams make ML repositories maintainable, testable, and release-safe — code, tests, and working systems, not strategy decks.

Get in touch: 📧 LinkedIn — describe your situation and I'll respond within a few days.

How to pay. Four things are clickable tiers on GitHub Sponsors — a 60-minute call, a workshop session, a two-session workshop package, and monthly priority work in my projects — with no purchase order and no invoice cycle. Everything else on this page is agreed by email first and then paid through the custom tier: repository reviews and CI work, Claude Code guardrails, longer or on-site workshops, retainers, sprint weeks, and any private-code delivery.

Two prices, one rule. The OSS price applies when the result is public: work on repositories under an OSI-approved licence, or an engagement where you agree to a public reference (your logo, a dated line in SPONSORS.md, a quotable sentence, and a short write-up I may publish — approved by you before it goes out). The standard price, roughly 50% higher, applies to private or proprietary code and to anything under NDA. The reference: yes / no question is settled in the scope email before work starts.

Scope limits. Sponsored priority work covers only the projects I own or co-maintain — AI-Rig, affordable-GPU-CI, fuse-augmentations, vision-delivery, pyDeprecate, cachier (co-maintained), and new projects of mine as they go public. The computer-vision libraries I maintain at Roboflow and the Lightning projects I co-created are funded elsewhere and are not for sale here. I don't sell merges or roadmap influence anywhere.

Typical engagement: 3–5 days for a report-only audit; 1–3 weeks for a sprint with working deliverables.

Practitioner sessions for engineering teams, built on your own repositories rather than slideware. This is the fastest way to move a whole team at once, and it is what most inbound requests ask for.

• T1 · GitHub platform & governance — rulesets, required checks, org policy, branch and tag protection, CODEOWNERS.
• T2 · GitHub Actions for polyglot estates — reusable workflows, matrices, caching, self-hosted and cheap GPU runners, cost control.
• T3 · Supply-chain security in CI — SHA pinning, token permissions, pull_request_target and pwn-request patterns, secret handling, scanning.
• T4 · Agentic engineering with Claude Code / Copilot — CLAUDE.md and rules, hooks, subagents, headless runs in CI, review guardrails, cost caps.

Every format includes a 30-minute scoping call, demos on your own repositories, slides, internal recording rights, and a written Q&A follow-up. Six formats, from a single 2-hour session to a four-session programme, priced from $1,000.

• Attendance up to 20 people is included; 21–40 is +25%; larger groups by quote — beyond 20 the demos stop working and the session turns into a lecture.
• Scheduling: single sessions within ~6 weeks, packages within ~8 weeks. One free reschedule if moved at least 14 days out; later changes forfeit 50%.
• Recording stays internal to your organisation. Publishing anything with your material needs written approval of the exact text.
• On-site (W5): EU destinations, 4 weeks' notice, travel and accommodation booked by you or reimbursed at cost, travel day billed at half rate.
• Programmes (W6) and anything above 40 attendees are scoped in writing before payment.
• Conference tutorials and meetup talks are arranged by email, not through this page.

Fixed prices for bounded work, floors for anything that needs scoping. Two columns: the OSS price when the result is public, the standard price for private or NDA work.

Prices are in USD, matching GitHub Sponsors. Only four of these are clickable tiers — the call, the single workshop session, the two-session package, and monthly priority work; the rest are agreed by email and paid through the custom tier, so the ladder stays short and nothing gets bought by accident. "From" figures are floors that a scope conversation turns into a quote.

What it covers

Repository structure, CI pipelines, release process, documentation gaps, issue/PR flow, contributor experience, and maintenance risk signals.

What you get

A prioritized findings report with concrete fixes — not a generic checklist. Each finding includes: the risk it represents, the fix, and estimated effort. Delivered as a PR or document you can act on immediately.

Who it's for

• Teams taking over an inherited or unmaintained repository
• Maintainers preparing a major release or community hand-off
• Sponsors or employers asking "is this project healthy enough to depend on?"

Price

Single-repo CI & Actions health check: $600 public / $900 private, agreed by email and paid through the custom tier, report within 10 business days. Full multi-repo audit: from $4,500 public / from $7,000 private, by quote.

What it covers

Versioning strategy, deprecation policy, migration warnings, compatibility layers, tests, and release notes for Python libraries undergoing a breaking change or long-term API evolution.

What you get

Working deprecation machinery (patterns from pyDeprecate), a migration guide for users, and a test matrix covering both the old and new API surface. Leaves your library with a documented deprecation process you can repeat on future changes.

Who it's for

• Library authors shipping a breaking v2 or v3
• Teams dealing with silent regressions from gradual API drift
• Maintainers who need to deprecate cleanly without breaking downstream users overnight

Price

Sprint week, scoped in writing: from $4,500 public / from $7,000 private, by quote and paid through the custom tier.

What it covers

Detection, segmentation, tracking, adapter, fixture, evaluation, and reproducibility gaps in CV systems built on detection/tracking stacks.

What you get

Identified gaps, added fixtures for critical adapter paths, reproducibility documentation, and a reliability report with actionable items ranked by risk to production.

Who it's for

• Teams shipping CV systems to production who need confidence their pipeline handles edge cases
• ML engineers inheriting a CV codebase with unclear test coverage

Price

Sprint week, scoped in writing: from $4,500 public / from $7,000 private, by quote and paid through the custom tier.

What it covers

.github defaults, AI-Rig-style review workflows, release checks, PR Evidence Cards, and human-in-the-loop review calibrated to your team's risk tolerance and repo size.

What you get

A configured .github overlay, working agent workflows for code review and release audit, a release checklist, and documented review processes your team can maintain. Based on patterns from .github and AI-Rig.

Who it's for

• Maintainers who want AI in their loop without losing control or shipping hallucinated reviews
• Teams scaling review capacity without scaling headcount
• OSS projects preparing for community growth who need consistent triage and release hygiene

Price

AI code review in your CI, one repo: $700 public / $1,000 private. Claude Code guardrails in a team repo: $900 public / $1,400 private. Both agreed by email and paid through the custom tier, delivered within 10–15 business days as a PR plus a handover call. Monorepos by quote. Prefer to train the team instead of having it built? See T4 under Workshops.

Thank you for supporting open source! 🙏