Own core pieces of the distributed stack that runs continuous training, inference, and real-time agent execution across our life-sciences platforms. Focus is reliability and hardware efficiency under production load.
Design and scale high-throughput inference engines and distributed serving runtimes.
Tune memory layout, GPU kernels, and inter-node communication.
Build and harden secure code-execution sandboxes for agent workflows.
Work directly with product engineers to keep frontier-model behavior stable in regulated environments.
Requirements
Systems programming (C++, Rust, or Python) plus performance-engineering experience.
Hands-on work with CUDA, Triton, Megatron-LM, SGLang, vLLM, or Ray.
Proven record building or maintaining large-scale ML infrastructure.
High ownership, fast iteration, and no tolerance for fragile shortcuts.
Frontier AI Red-Team & Eval Collaborator
June 18, 2026 · Closed · Short-Term / Consulting (1–3 Months) · $1,200 – $1,600 / day · Remote / Hybrid (Dallas, TX or Budapest)
Role & Scope
Short project to surface subtle failure modes in frontier models before they reach production life-sciences workflows. Emphasis on domain-specific degradation, reward hacking, and edge-case safety.
Build lightweight eval harnesses and targeted adversarial prompts.
Stress-test sycophancy, tool-use failures, and domain drift.
Produce concise technical write-ups, risk notes, and reusable benchmark suites.
Sit in small engineering reviews that turn findings into concrete guardrail changes.
Requirements
Prior research or applied work in model evaluation, red-teaming, or RLHF mechanics.
Direct experience with prompt safety, degradation tracking, or custom eval design.
Ability to turn raw behavioral anomalies into clear, reproducible reports.
Self-directed and able to deliver on a tight timeline.
Turn internal technical detail into precise external material for life-sciences executives and engineering buyers. No marketing gloss — just accurate explanations of platforms, pipelines, and clinical infrastructure.
Write technical articles, architecture notes, product overviews, and case studies.
Extract exact domain detail from ML and compliance teams.
Repurpose source documentation into multi-channel pieces.
Set and enforce writing standards, terminology, and basic performance tracking.
Requirements
3+ years editorial or technical content work in enterprise SaaS, AI, or life sciences.
Portfolio that shows clean, accurate explanation of complex systems.
Strong narrative sense paired with hands-on execution.
Strict preference for technical accuracy over promotional language.