Ellwood Consulting Hiring! Full Time Machine Learning Specialist in - Ricebowl

Technical Lead, Machine Learning

The role

We need someone who can take research ideas and turn them into ML systems that actually work in production — reliably, at scale, day after day. You'll sit right at the crossroads of research, infrastructure, and product, making sure models don't just perform well in a notebook but hold up under real users, real load, and real constraints.

What you'll be doing

• Owning ML systems end-to-end — data pipelines, training workflows, evals, inference, deployment. The whole pipe, not just one slice of it.
• Fine-tuning and adapting models with modern techniques (LoRA, QLoRA, SFT, DPO, distillation) — whatever gets the best result for the constraints we're working with.
• Building inference systems that don't force you to choose between fast, cheap, and reliable — or at least finding the right trade-off when you do.
• Keeping our training data (synthetic and real) clean, well-structured, and useful.
• Building evaluation pipelines that check not just "does it perform well" but "is it safe, robust, and fair" — working closely with research leadership on this.
• Getting models into production and keeping them there: GPU optimization, memory efficiency, latency, scaling — the unglamorous work that makes everything else possible.
• Working shoulder-to-shoulder with application engineers so ML doesn't feel bolted on to our backend, mobile, and desktop products — it feels native.
• Making smart, pragmatic calls and shipping fast, then learning from how things actually get used in the wild.

What success looks like

• Research turns into shipped, production-ready work — with real targets, not vague hopes.
• Our pipelines and inference systems are stable and easy to maintain — not held together with duct tape.
• When something breaks in production, it gets caught and fixed fast, before users really feel it.
• The people around you feel supported and unblocked, not stuck waiting on you.
• Every iteration on our models makes the product measurably better and safer over time.

Stack

Python, PyTorch/JAX, GPU-based training and inference systems.

Who you probably are

• You've shipped ML systems real people use — not just impressive demos.
• You know large models well enough to know how and where they break.
• You write code you're proud of, and you care about correctness, not just "it works on my machine."
• You're self-directed and pragmatic — you take ownership of outcomes, not just tasks.
• You communicate clearly and work well in a small team where trust matters more than process.