Cégleírás / Organisation/Department
Our client is a US-headquartered, market-leading enterprise with a multi-decade history and an agile, global network. We are seeking a Senior Data Scientist to join their expanding Budapest team as a senior technical individual contributor — someone who builds models that go to production, not models that live in notebooks.
...
Situated within the Data Engineering organization alongside ML engineers, data engineers, and data architects, you will own the applied science behind predictive and agentic AI capabilities — from problem framing and experimentation through production validation and ongoing performance ownership. Formulating problems rigorously and iterating quickly, you will partner directly with ML engineers to ensure every model is production-viable from day one.
Pozíció leírása / Job description
- Own the full applied science lifecycle for predictive and generative AI use cases - problem framing, data exploration, feature discovery, model development, offline evaluation, production validation, and post-deployment performance ownership
- Design and execute experiments with production constraints in mind from the start - latency budgets, serving costs, data availability, retraining cadence, and failure modes are part of your design process, not afterthoughts handed off to engineering
- Develop predictive models (classification, regression, forecasting, ranking, anomaly detection, recommendation) using modern frameworks (PyTorch, XGBoost, LightGBM, scikit-learn) and deliver them in production-ready form through MLflow, Databricks Model Serving, or equivalent infrastructure
- Build and refine applied LLM solutions — RAG architectures, fine-tuned domain models, prompt engineering pipelines, structured output extraction, and evaluation harnesses — with a focus on reliability, cost control, and measurable quality at production scale
- Design evaluation frameworks that go beyond offline metrics — A/B testing strategies, online performance tracking, business KPI attribution, and model degradation detection that give stakeholders and engineering teams a shared, honest view of model value
- Partner directly with ML engineers on productionalization, contributing clean, well-tested, well-documented code rather than handing off research artifacts that require translation
- Collaborate with data engineers on feature engineering, data quality requirements, and pipeline design, defining upstream data contracts that your models depend on
- Work with product and business stakeholders to frame ambiguous business problems into tractable data science formulations, pushing back when a simpler analytical or rule- based approach outperforms a model
- Conduct rigorous exploratory data analysis to uncover data quality issues, distribution shifts, leakage risks, and feature viability before investing in model development
- Mentor junior data scientists and analysts on applied methodology, experimental rigor, and the discipline of building for production rather than demonstration
Elvárások / Requirements
- 6+ years in applied data science or machine learning roles with recent, demonstrated production delivery — models you built are running in production today, serving real decisions or real users
- Deep proficiency in Python and the modern applied data science stack — PyTorch, scikit-learn, XGBoost/LightGBM, Pandas, Polars, NumPy, and SQL as a first-class analysis tool, not an afterthought
- Production experience with the Databricks ecosystem — notebooks for development, MLflow for experiment tracking and model registry, Feature Store for feature management, Unity Catalog for data and model governance, and Databricks SQL for analytical workloads. Working knowledge of supporting AWS services (S3, SageMaker, Bedrock, Athena)
- Hands-on production experience with LLM-powered applications — RAG pipeline design, embedding model selection and optimization, prompt engineering, fine-tuning (LoRA, QLoRA, or full), and structured evaluation of generative outputs beyond vibes- based assessment
- Demonstrated experience with agentic AI patterns in applied contexts — building or validating agent workflows where a data scientist's role includes defining success metrics, designing evaluation harnesses, and tuning system behavior through prompt architecture and retrieval optimization
- Strong statistical foundations applied in practice — experimental design, causal inference, hypothesis testing, and Bayesian reasoning used to make real decisions, not just listed on a resume
- Experience with modern MLOps workflows — model versioning, automated retraining triggers, CI/CD for model artifacts, feature/target drift monitoring, and shadow deployment or canary release patterns
- Track record of delivering measurable business impact through applied models with the ability to articulate that impact in terms stakeholders understand — revenue, cost reduction, conversion, retention, not just AUC and F1
Preferred Qualifications
- Experience with real-time or near-real-time inference use cases — online feature computation, streaming prediction, and the latency/throughput tradeoffs involved in serving models at scale
- Production experience with advanced retrieval systems — hybrid search (dense + sparse), re-ranking models, query decomposition, and evaluation of retrieval quality as a component of end-to-end LLM system performance
- Hands-on experience with model fine-tuning and alignment techniques (RLHF, DPO, ORPO) applied to domain-specific use cases with measurable quality improvements over base models
- Familiarity with causal ML methods (uplift modeling, causal forests, double ML) applied to business problems like treatment optimization, pricing, or personalization
- Experience building and maintaining evaluation datasets, annotation pipelines, and ground-truth collection processes that enable ongoing model quality measurement
- MS or PhD in a quantitative field — statistics, computer science, applied mathematics, physics, economics, or equivalent (practical production experience weighted equally with academic credentials)
- Fluency in Hungarian is essential.
Amit kínálunk / Offer
- High Technical Impact: A unique professional challenge to shape production predictive AI models and agentic orchestration workflows for a stable, market-leading global enterprise.
- Competitive Compensation: Senior level base salary and annual target bonus.
- Benefits Package: Cafeteria allowance and private health insurance package.
- Flexibility: Hybrid working model requiring 2 days of office presence (Budapest) and offering 3 days of Home Office per week.
Kapcsolattartó / Information
Ildikó Mező-Mészáros - ildiko.mezo-meszaros@randstad.hu
Réka Lukáts - reka.lukats@randstad.hu