SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 250,000 - 450,000 / annual
Vizcom is a Series B design platform used by professional design teams at Nike, GM, New Balance, and Hasbro for sketching, rendering, color/material work, 3D modeling, and export. The company has raised $52M from top-tier investors including Radical Ventures and Index Ventures.
You will be the second research engineer on a two-person research team, working alongside the engineer who built Vizcom's current post-training stack. Your mission: build models that understand the judgment designers apply when moving work from pencil to product. This is a research role deeply coupled to product—your models ship to working professionals within weeks, not quarters.
You will own the post-training roadmap end-to-end: research planning, reward and preference modeling on five years of professional design decisions, the full method stack (supervised fine-tuning, distillation, reinforcement learning), research standards and evaluation bars, and the feedback loop with product. The company has accumulated a research log documenting every dead end already explored, giving you a head start.
The core challenge: a designer's choice among candidates entangles the generator's style, the moment in the process, and the designer's intent. Separating these signals requires both clever objectives and rigorous engineering—correct training code, honest evaluations, and pipelines that don't lie.
In your first 90 days, you'll immerse in the data and existing research log (days 1–30), produce one result that survives the audit bar on historical data (days 30–60), and write the roadmap from historical signal to a closed loop (days 60–90).
You should have strong programming skills, hands-on experience post-training diffusion or flow models (supervised fine-tuning, preference optimization, or RL), and genuine excitement about product-coupled research with audit-grade rigor. Nice-to-haves include high-performance training/inference implementations and experience making physical things or deep interest in the people who do. The company values people more interested in how professionals decide than in what the internet likes.
Visa sponsorship is available. The company operates on research-log culture where negative results are celebrated, nothing scales on vibes, and hiring happens through paid work trials on real problems.
