Machine Learning Ops Expert
Machine Learning Ops Expert
雇用形態:
正社員
職種:
データ分析
言語:
English Level - Advanced (TOEIC 860),Japanese Level - None
勤務地:
Minato-ku
給与:
¥8,000,000.00 - ¥30,000,000.00 年収
求人ID:
494263
A cutting-edge Healthtech start-up is looking for a Machine Learning Ops Expert - LLM Fine-Tuning & Production Deployment (Scientific Journals).
Responsibilities:
- Design and build end-to-end multimodal document AI pipelines to extract complex structured data (tables, figures, charts, captions, images) from scientific papers and PDFs
- Fine-tune domain-specific LLMs (e.g., LLaMA/GPT-class) using scientific journal corpora with techniques such as SFT, LoRA/QLoRA, and related methods
- Develop and manage full MLOps workflows (data ingestion → training → evaluation → deployment) for production-grade LLM systems
- Deploy and operate scalable inference APIs in production environments with CI/CD, Docker/Kubernetes, and cloud infrastructure (AWS/GCP/Azure)
- Establish robust evaluation frameworks including benchmark design, regression testing, hallucination control, and human-in-the-loop validation
- Ensure reliability and performance in production through monitoring, observability, drift detection, automated retraining, and cost/latency optimization
Required skills:
- Strong experience with scientific paper document AI / multimodal extraction — extracting structured data from PDFs and scientific journals, including tables, figures, charts/plots, captions, images/photos, OCR + layout parsing pipelines
- Proven experience fine-tuning LLMs (e.g., LLaMA/GPT-class) on domain-specific corpora (scientific journals, complex structured text)
- Proven track record building end-to-end MLOps pipelines (data → training → evaluation → deployment) and deploying LLMs to production (scalable inference APIs, CI/CD, reliability, monitoring, rollback)
- Strong understanding of LLM architecture + training methods (SFT, LoRA/QLoRA, RLHF/DPO), tokenization, context limits, hallucination control
- Strong knowledge of LLM evaluation: task-specific metrics, benchmark design, error analysis, regression testing, human-in-the-loop evaluation
- Experience operating models in production: observability, drift detection, automated retraining, cost/latency optimization
- Hands-on with modern stack: Python, PyTorch, Hugging Face/Transformers, Docker/Kubernetes, cloud (AWS/GCP/Azure)
Why should you apply:
- Join a company with a bold mission to elevate humanity through the fusion of AI, biotechnology, and human augmentation
- Receive near-zero-cost, founding-level stock options with significant upside potential. The company also offers liquidity opportunities at each funding round — a rare and valuable benefit in early-stage startups
- Engage in cross-disciplinary innovation involving scRNA-seq, spatial transcriptomics, PBPK/PD modeling, large-scale ML/DL, and chatbot development. This is a unique opportunity to grow your expertise across multiple frontier technologies
Company Overview:
A HealthTech, Biotech firm focused on leveraging clinical pharmaceutical data and big data to revolutionize drug development and health monitoring. Its focus is on a build an easy-to-use chatbot that can perform complex large quantitative modelling. Specifically, it will develop a next-generation PBPK/PD simulator that combines massive scientific & mathematical datasets with machine learning, deep learning, and natural language processing. The firm has already formed research partnership with Japanese university and labs and is targetting both Japan and US market, which is why frequent travel to US is expected. The CEO has a strong engineering mindset, so the working style is highly flexible — including options for remote work, flexible hours, and minimal constraints.
Working Hours: 9:00 – 18:00 (Mon-Fri) / Fully Flex Time
Working Style: Work in office / Remote work *flexible
Holidays: Saturday, Sunday, and National Holidays, Year-end and New Year Holidays, Paid Holidays.
Services/Benefits: Social Insurance, Founding-level stock options (near-zero cost) with high upside potential, Stock monetization opportunities at each funding round, Comprehensive relocation support to Japan (visa, moving, housing, tax support, bilingual staff assistance)
Interview Process : 2-4 times
シェア
