Lead Model Accuracy Development and Test Engineer (Datacentre AI Engineering) - Riyadh, KSA

Company:Qualcomm Middle East Information Technology Company LLCJob Area:Engineering Group, Engineering Group > Software EngineeringAbout the RoleWe are seeking a hands‑on technical leader to own end‑to‑end model accuracy for large‑scale deep learning workloads running on data‑center accelerators. As the Lead for Model Accuracy Development & Test, you will set the technical direction, mentor engineers, and drive cross‑functional programs to define KPIs, build accuracy pipelines, debug failures, and deliver accuracy parity across frameworks (PyTorch/ONNX) and hardware targets (AI100/AI200, NVIDIA, Gaudi, TPU). You will combine deep systems knowledge with pragmatic program execution to unblock production deployments and raise the bar for accuracy engineering.Key ResponsibilitiesTechnical Leadership & StrategyDefine the multi‑quarter accuracy roadmap, KPIs, and bar for productization across precision modes (FP32/FP16/INT8/FP8/INT4)Own architecture for accuracy evaluation pipelines, including data prep, replay, metrics, dashboards, and automated triage.Make design decisions on quantization/calibration strategies, operator implementations, and accuracy‑preserving optimizations (TensorRT, ONNX Runtime, AITemplate, Triton).Delivery & Cross‑Functional OwnershipDrive cross‑team execution with model onboarding, runtime, compiler, firmware, and HW perf teams to reach accuracy sign‑off.Establish SLAs and acceptance criteria for accuracy parity across PyTorch → ONNX → backend engines and across hardware.Create clear investigation plans for deltas (pre/post‑processing drift, tokenization mismatch, QDQ placement, operator fallback).Debugging & AnalysisLead deep‑dives for failure triage using intermediate dumps, layer‑wise comparisons, attribution tools, and sensitivity analysis.Perform slice‑based analysis (batch, concurrency, sequence length, domain shifts) and design experiments for recovery (calibration, fine‑tuning, hyperparameter sweeps).Institutionalize patterns/playbooks for recurring issues (normalization overflow, activation scaling, precision loss, layout drift).People Leadership & MentoringMentor 3–6 engineers; conduct design reviews, code reviews, and guide experiment methodology and result interpretation.Partner with recruiting; define interview rubrics; onboard and grow the accuracy engineering discipline within the team.Quality, Compliance & DocumentationEnsure reproducibility and auditability of results (versioning of datasets, artifacts, firmware/runtime).Publish clear dashboards, RFCs, and decision logs; drive stakeholder communication and status.Required Experience & Skills10+ years of industry experience in AI/ML, with significant ownership of model accuracy and evaluation.Expertise in architectures (Transformers, CNNs, Diffusion, MoE) and their accuracy sensitivities.Proven depth with inference runtimes/backends (TensorRT, ONNX Runtime, AITemplate, Triton) and graph conversion (PyTorch → ONNX → engines).Strong quantization background (INT8/FP8/INT4), calibration methods, QAT, and mixed‑precision workflows.Hands‑on engineering in Python; solid software engineering practices (testing, CI, packaging).Experience comparing results across hardware/software stacks (AI100/AI200, NVIDIA A100/H100/B200, Gaudi, TPU) and firmware/runtime versions.Fluency with accuracy metrics, statistical analysis, visualization, and experiment design.Demonstrated ability to lead cross‑functional programs and mentor engineers.We’d love to seeExperience with video generation/I2V accuracy and multi‑modal benchmarking.Familiarity with LLM/VLM accuracy tooling (lm‑eval, HELM) and dataset curation.Background in distributed systems/Kubernetes and cloud inference services.QualificationsBS/MS in Computer Science, Electrical/Computer Engineering, or related field; PhD is a plus.10+ years of relevant software/ML experience; 3+ years in a leading/mentoring capacity.Strong problem‑solving and communication skills; ability to influence across organizations.Success Metrics (Examples)Accuracy parity vs. PyTorch/ONNX baselines within target MAD/PSNR/SSIM thresholds for each model + precision mode.Time‑to‑resolution for P0 accuracy regressions (median/95th).Automation coverage of accuracy evaluation across supported frameworks and hardware targets.Mentorship outcomes: onboarding time reduction and quality of design reviews.What's on OfferSalary including housing & transport allowanceStock (RSU's) and performance related bonus16 weeks fully paid Maternity Leave6 weeks fully paid Paternity LeaveEmployee stock purchase schemeChild Education AllowanceRelocation and immigration support (if needed)Life and Medical InsuranceLive+ Well Reimbursement for health and recreational membership feesMinimum Qualifications• Bachelor's degree in Engineering, Information Systems, Computer Science, or related field and 4+ years of Software Engineering or related work experience.• Master’s degree in Engineering, Information Systems, Computer Science, or related field