AI Engineer: Post-Training

Overview

In this role, you own the product capability to improve LLMs for PhysicsX and its customers. You will work closely with Research to turn post-training methods into a production-ready solution, starting with strong open-model performance on our benchmarks. The role aims to enable customers to train models on their own workflows while keeping data and IP on their platforms. You will manage the pathway from base models to deployable improvements and shape a scalable platform. This is an opportunity to impact how industrial AI is applied across aerospace, automotive, and manufacturing domains.

Responsibilities

• Own the path from a base model and customer workflow to a measured, deployable model improvement
• Build reliable post-training harnesses covering data preparation, training, evaluation, model artifacts and hand-off to serving
• Improve open models against PhysicsX’s agent and engineering benchmarks, using evidence to decide training versus prompt/tools changes
• Develop and operate methods spanning SFT, preference optimisation, RLHF/RLAIF, RL with verifiable rewards, distillation, rejection sampling
• Build reproducible synthetic-data and expert-data pipelines with quality filtering, provenance, licensing and contamination controls
• Design reward signals, verifiers and rollout workflows while detecting reward hacking, benchmark overfitting, model collapse and unintended regressions
• Make post-training available through managed workflows and expert-facing interfaces
• Ensure customer data and artifacts remain within the customer’s platform across SaaS, VPC and air-gapped deployments
• Work on cross-product and public benchmarks; own experiment diagnostics with Research and aligned approvals
• Collaborate with LLM Self-Hosting & Model Routing Lead to move trained models into production serving and routing
• Set the product roadmap for model improvement balancing research opportunity with customer value and production timelines
• Help determine how this capability should grow and influence team structure

Key requirements

• Measurably improved model behaviour through post-training and production-ready capability
• Hands-on across data, training, evaluation and debugging; able to take an experiment from hypothesis to deployed artifact
• Solid understanding of modern post-training methods and how to select them by task and evidence
• Experience building high-quality training datasets from expert, behavioural or synthetic sources
• Ability to diagnose training and behavioural failures (instability, drift, leakage, overfitting to evals)
• Product-minded: start from user workflow, define success, make trade-offs, enable operation by others
• Experience as accountable product owner of a technical capability while collaborating with researchers and platform teams
• Strong Python and PyTorch or equivalent frameworks
• Distributed training, accelerator utilisation, checkpointing, reproducibility and experiment tracking
• Data governance, privacy, provenance and safe handling of customer data and artifacts
• Model packaging and the path from training to production serving
• Clear technical writing and cross-team communication

• Product-minded
• Collaborative
• Analytical
• Python
• PyTorch
• Distributed training

…