Crypto Data Scientist Machine Learning - LLM Engineer (Global - Remote - Non.US)

�

� Description

• • Shape the future of AI-driven crypto investing. As Token Metrics’ next Machine Learning – LLM Engineer you will own the end-to-end lifecycle of models that turn billions of on-chain and off-chain data points into actionable insights for investors in 50+ countries. Your work will directly influence portfolio construction, risk scoring, and price-prediction engines used by retail traders and institutional funds alike.
• • Partner with the Head of Data Science to translate business goals into ML objectives. You will scope problems such as “How do we predict a token’s 7-day volatility from Telegram sentiment, GitHub commits, and DEX liquidity curves?” and design experiments that move key KPIs like Sharpe ratio uplift and signal latency.
• • Architect and productionize self-running AI pipelines. Build modular Python services that ingest streaming market data, engineer features in real time, train gradient-boosted and transformer models, and push low-latency predictions to our API layer. Ensure every component is containerized, monitored, and horizontally scalable on AWS/GCP.
• • Fine-tune large language models for crypto domain tasks. Curate specialized corpora (whitepapers, governance forums, Twitter threads) and use LoRA/QLoRA to adapt open-source LLMs for sentiment extraction, summarization, and question-answering. Implement LLM observability with prompt drift detection and hallucination scoring.
• • Stress-test models across market regimes. Simulate black-swan events, exchange outages, and liquidity shocks to measure robustness. Present statistical significance tests and confidence intervals that give traders clear guidance on when to trust—or override—model outputs.
• • Optimize existing libraries and frameworks. Profile training jobs that crunch terabytes of candlestick data; refactor code to cut GPU hours by 30 %. Contribute improvements back to open-source projects like PyTorch Lightning and Hugging Face Transformers.
• • Collaborate cross-functionally. Work with quant researchers to encode new alpha factors, with backend engineers to expose gRPC endpoints, and with product managers to design human-in-the-loop review workflows. Translate complex findings into crisp visualizations for non-technical stakeholders.
• • Champion MLOps best practices. Implement CI/CD pipelines that run unit tests, data-quality checks, and model-validation suites on every pull request. Maintain feature stores and model registries with full lineage and rollback capabilities.
• • Stay ahead of the curve. Run weekly paper-reading sessions on arXiv preprints, DeFi governance proposals, and new tokenomics designs. Rapidly prototype promising ideas in Jupyter notebooks and pitch the top 10 % for full-scale development.
• • Document everything. Produce clear READMEs, architecture diagrams, and runbooks that allow any team member to reproduce experiments or debug incidents at 3 a.m. without paging you.
• • Mentor junior data scientists. Conduct code reviews, pair program on tricky tensor manipulations, and share battle-tested tips for wrangling messy on-chain datasets. Help grow the team’s collective expertise in both machine learning and crypto fundamentals.
• • Drive measurable impact. Within your first 90 days you will ship a fine-tuned LLM that improves sentiment signal F1 score by 15 %, and within six months you will lead a project that increases annual recurring revenue by enhancing prediction accuracy for our top-tier institutional clients.

� Requirements

• • Bachelor’s degree in Computer Science, Data Science, Mathematics, or related field; Master’s in computational linguistics, data analytics, or similar strongly preferred
• • 2+ years of hands-on experience as a machine-learning engineer delivering models to production
• • Advanced proficiency in Python (NumPy, pandas, PyTorch, scikit-learn) and working knowledge of Java or R
• • Demonstrated experience fine-tuning large language models (LoRA, RLHF) and implementing LLM observability tooling
• • Solid grasp of statistics, linear algebra, and algorithm design; familiarity with crypto or web3 data sets is a plus

️ Benefits

• • Fully remote, asynchronous-first culture—work from Tirana or anywhere outside the U.S. with flexible hours
• • Competitive global salary paid in stablecoins or fiat, plus performance token bonuses tied to model profitability
• • Annual learning stipend (courses, conferences, GPUs) and dedicated time for open-source contributions