6 days ago
Remote, SpainSenior
Responsibilities
- Implement and evaluate targeted open-source LLM adaptations for financial-domain use cases.
- Prototype preference-alignment approaches including DPO and PPO.
- Benchmark and optimize model serving for low latency and high throughput.
- Apply model quantization and VRAM optimization using Triton Inference Server, TEI, and vLLM.
- Prototype advanced search and retrieval methods, including Matryoshka embeddings, late interaction models, and hybrid search pipelines.
- Package training, evaluation, and serving scripts into clean, reproducible deliverables using AWS SageMaker and Docker.
- Build synthetic-data generation and automated evaluation harnesses to measure cost, latency, and quality trade-offs.
- Deliver autonomous, well-documented, modular code and rapidly validate ideas through empirical testing.
Requirements
- Master's or PhD in Computer Science, Machine Learning, or a quantitative field, or equivalent practical experience.
- Proven experience delivering production-grade ML models and proof-of-concepts in Search, Information Retrieval, or LLM infrastructure.
- Deep hands-on experience with Python, PyTorch, and the Hugging Face ecosystem, including Transformers, PEFT, and Accelerate.
- Practical experience with model quantization, VRAM optimization, and high-performance model serving.
- Experience using AWS SageMaker for training and deployment, plus experiment tracking and tracing tools such as MLflow and Opik.
- Fluent written and verbal English communication skills.
- European legal working status and alignment with the EU time zone.
- Experience with RLHF or Direct Preference Optimization is a bonus.
- Familiarity with financial market data and financial text processing is a bonus.
Benefits
- Initial three-to-six-month independent-contractor engagement with potential extension based on project milestones and results.
- Embedded independent-contributor role with the internal Search & Recommendation engineering team.
- Competitive daily or project-based contract rate commensurate with experience.
- European legal working status and EU time-zone alignment are required.
- Equal opportunity employer committed to diversity.
Categories
About RavenPack
RavenPack builds NLP-driven analytics that turn unstructured content—news, filings, and social media—into machine-readable data for quantitative investors and risk teams. It sells data feeds, APIs, and a SaaS platform, including Bigdata.com, an agentic AI workspace launched in 2024 for financial research. Founded in 2003 and headquartered in Marbella, Spain, the privately held firm is used by hedge funds, banks, and asset managers.