AI Utilization & Data Analysis Engineer (Cyber ​​Defense - Gov't Project)

Description

Solafune Co., Ltd. is a startup aiming to build a "Planetary Intelligence OS." With satellite data analysis and AI analysis at its core, it is rapidly building a track record of transactions with government agencies in fields such as defense and security, infrastructure monitoring, and environmental monitoring.

In a security system optimization project for government agencies, this position utilizes local LLMs and statistical analysis to optimize security parameters. In a highly confidential, isolated environment where external cloud AI cannot be used, you will utilize locally deployed open-source LLMs to comprehensively analyze verification result data and external interview information to derive optimal parameter candidates. We welcome engineers who want to take on the challenge of "using AI in a closed environment."

Main Job Responsibilities

• Construction and operation of a local LLM environment (without using external services): Deploying models on GPU servers using Ollama
• Implementation of parameter optimization loops: Inputting verification result data for the LLM to generate, evaluate, and narrow down parameter candidates
• Structured organization and pattern analysis of external interview information and existing case data
• Implementation and improvement of analysis pipelines using Python/LangChain
• Preparation of analysis result reports (utilized as quantitative basis for design policy documents)
• Construction and maintenance of an analysis environment in an isolated (network-disconnected) environment

Requirements

• Python 3.x: Data processing, analysis, REST API implementation (Pandas / NumPy / FastAPI, etc.)
• Practical experience in machine learning and statistical analysis: Any of parameter optimization, anomaly detection, time-series analysis, etc.
• Experience utilizing local LLMs or open-source LLMs: Ollama / llama.cpp / Hugging Face, etc.
• Experience using data visualization tools: Jupyter Lab / Matplotlib / Plotly, etc.
• Basic operation of Linux servers and GPU environments (CUDA)
• Prompt engineering capability: Ability to design inputs that LLMs can effectively utilize based on technical specification text
• Experiment management capability: Ability to organize conditions and results for each iteration to ensure reproducibility
• Resourcefulness in a closed environment: Ability to maintain the quality of analysis even under constraints where external services cannot be used

Tools and Product Experience

• Local LLM: Ollama + Llama3/Mistral or llama.cpp + GGUF model (Experience with either one is acceptable)
• LLM Framework: LangChain or LlamaIndex (Experience with either one is acceptable)
• Optimization: Optuna or scipy.optimize (Experience with either one is acceptable)
• Container: Docker (Basic operation)

Preferred Experiences

• Machine learning-related qualifications: G Certificate, E Certificate, AWS ML Specialty, etc.
• CompTIA Security+ (Recommended as a guarantee of basic security knowledge)

We are looking for

• Those who can understand the characteristics of projects for government agencies and highly confidential projects, and approach their work with a strong sense of ethics and responsibility.
• Those who prioritize information security and compliance, and can thoroughly manage confidential information appropriately.
• Those who can adhere to internal and external rules and security policies, and carry out their duties carefully and honestly.
• Those who can flexibly accommodate regular business trips to designated facilities in Tokyo.
• Those who have no resistance to working in a secure environment isolated from networks and can focus on their work.
• Those who can collaborate smoothly with stakeholders and contribute to generating results as a team.
• Those who feel fulfillment in highly public missions and have an interest in social infrastructure and defense/security fields.