ComfyUI Engineer.
Own the workflows that power millions of AI images every month. You think in nodes, you debug in graphs, you care obsessively about output quality.
Runflow is the AI image and video infrastructure layer for production. We power the AI headshots, fashion shoots, and product photography that companies like BetterPic and BetterStudio ship every day, millions of images a month, with quality scored before delivery.
We're hiring a ComfyUI Engineer to own the workflows that produce them. This isn't a research role. You'll ship production workflows that directly impact output quality, customer retention, and revenue. You're the person who decides which checkpoints, which LoRAs, which control nets, which sampler, and how it all stays under VRAM and latency budgets at scale.
If you debug node graphs in your sleep, train LoRAs for fun, and have strong opinions on face similarity vs. likeness preservation, this is for you.
Quick facts
What you'd ship in your first 6 months.
Concrete, shippable, on the live roadmap. Not hypothetical. We expect every one of these to land.
Ship a face-similarity workflow that holds 99%+ likeness across 500-image batches
Pick the checkpoints, design the conditioning chain, calibrate the IP-Adapter / face-swap stack, and run the eval. Hand it off to Sentinel for live scoring.
Cut p95 latency on the BetterStudio pipeline by 40%
Profile VRAM, swap samplers, prune the graph, parallelise where it pays. The win has to hold across realistic batch shapes, not synthetic benchmarks.
Train a domain-specific LoRA from scratch for a new product vertical
Curate the dataset, set up the training loop, ship the LoRA into the workflow, validate against held-out evals. Document the recipe so the next vertical takes a week instead of a month.
What you'll be doing.
You'll own the full lifecycle, from prototype to production at scale.
Build, optimise, and debug ComfyUI workflows for human-centric and product image generation
Profile node chains. Cut VRAM, latency, and cost without sacrificing quality
Wire workflows into Runflow's Solutions API, Python, TypeScript, REST
Run batch inference and post-processing at production scale (millions/mo)
Iterate on output quality. Hand off failure modes to Sentinel; learn from its scores
Build dynamic conditioning systems, variable poses, styles, backgrounds, products
Maintain a versioned, documented workflow repo other engineers can read
Train and fine-tune checkpoints + LoRAs (Flux, SDXL, Wan-style pipelines)
About you.
You may be a good fit if…
- You've shipped ComfyUI in production for 6+ months, setup, custom nodes, debugging gnarly graphs
- You have deep working knowledge of diffusion models (SDXL, Flux, the open-source release of the month)
- You've built workflows tuned for realism and identity preservation, face similarity is second nature
- You're comfortable with ControlNets, IP-Adapters, segmentation, and the full conditioning toolbox
- You optimise compute. You know when to swap a sampler vs. retrain a LoRA
- You're fluent in GitHub, Python, and REST, the workflow is just one part of a system
Strong candidates also have…
- Have trained LoRAs or fine-tuned checkpoints (Flux/SDXL/Wan) end to end
- Use the diffusers library and Hugging Face tooling fluently
- Have built integrations connecting ComfyUI to external services (Plytix, Shopify, custom CMSes)
- Have shipped production pipelines at scale (>100K images/day) and lived to tell the war stories
- Have a public portfolio of workflows you'd be proud to walk us through
What we're not looking for
A specific number of years. A specific degree. A specific stack. We hire on whether you can ship end-to-end and whether you have taste. Everything else is noise.
What we work with.
Backend
Workflow
Infra
Tools
Five steps. Decision speed is part of the offer.
We move fast because senior candidates with multiple offers reward the team that respects their time. Whole loop fits in two weeks.
Application submitted
Form, ~5 minWe read every word of every application. No silent rejections, ever.
Take-home challenge
2 hours max, your timeA small, real Runflow problem. We score the prompts and decisions you made as much as the output itself.
30 minutes with the CTO
30 min, liveQuick conversation about your take-home, the team, and how you work day-to-day.
Paid challenge, 2 days
2 days, we pay for your timeA bigger problem we pay you to work through. We care about outcomes, not process. AI tools more than welcome.
Closing interview
90 min, liveCasual chat with the two founders.
Show, don't tell.
We value proof over promises. When you apply, include examples. Things that stand out:
Ready to ship?
Five minutes. We read every word. Yes / no / not-now within 5 business days, always.
Maybe one of these instead.
We hire builders, not roles. If none fit exactly, the open application is at the bottom.