Here is the formatted text:
Machine Learning Engineerr | (TokyoDev)
The Problem
Job matching is still keyword-based. "Python engineer" "Python job." Were building something fundamentally different: AI that matches people based on values, aspirations, and behavioral fit, not just skills. Our system explains why someone would thrive at a company andwhy a company should hire them-not just that requirements are met. Not recommendations. Meaning-based matching.
We have 200 paying companies and thousands of active users. The matching works. Now we need to make it exceptional.
What Youll Build
### 1. Memory Layer for Continuous Learning
Users have conversations with our AI over weeks and months. Youll build a system that:
- Extracts meaningful information from unstructured conversations (goals, values, preferences, work history)
- Stores both structured facts and semantic understanding
- Retrieves relevant context in <100ms to power personalized matching
- Handles evolution: People change jobs, preferences shift-the system should track this
Challenge: This is novel territory. Memory extraction from conversational AI at scale hasnt been solved. Youll design it from scratch.
### 2. Explainable Matching Engine
Generate recommendations with real explanations: "Company X fits because they prioritize remote work (you mentioned this last week) and have a strong mentorship culture (aligns with your leadership style)."
Challenge: Make LLM explanations trustworthy. No hallucinations, grounded in actual data, consistent across users.
### 3. Hybrid Matching Architecture
Design and implement matching that combines LLMs, rule-based logic, and traditional ML where each excels. Not everything needs a neural net.
Challenge: Know when to use deterministic rules vs. learned models. Optimize for explainability and cost, not just accuracy.
### 4. Company Knowledge Structures
Build semantic representations of company information-business model, culture, hiring requirements, and implicit knowledge that doesnt appear in job descriptions.
Challenge: Scale from 200 partner companies to 1,500+ with quality data sourcing, fast retrieval, and cost optimization.
### 5. Training Data & Evaluation
Design what to label, how to label it, and what "good" means. Build evaluation frameworks for subjective quality (Is this match good? Is this explanation helpful?). Run A/B tests. Benchmark models.
Challenge: Extract value from small datasets. You have 200 companies, not 200,000-design learning strategies that work with limited examples.
How Youll Work
- Cross-functional collaboration: Youll work directly with Product Managers and Business Development in rapid hypothesis implementation verification cycles.
- Two-sided optimization: Your models must satisfy both job seekers AND hiring companies. User satisfaction and company hiring decisions are both success metrics.
- LLM strategy design: Beyond calling APIs, youll design prompt strategies and learning approaches based on inference results. Meta-level thinking about how to use LLMs effectively.
Welcome Skills
- Total ownership: You define the ML vision, architecture, and roadmap
- Build the team: Hire and lead future ML engineers as we scale
- Founding engineer impact: Significant equity and organizational influence
- Solo execution initially: Can self-direct without needing ML peers (for now)
Reality check:
- Youll be the only ML engineer for ~6 months until we hire #2
- No ML peers to review your code or validate decisions
- CTO provides infrastructure support but isnt an ML specialist
- Small data environment: 200 companies, not 200,000
- Two-sided marketplace: optimizing for users AND companies
- This is high autonomy + high responsibility
- If you thrive with ML collaboration, this isnt the right fit yet
Application Overview
- Salary Range: ¥10,000,000 to ¥15,000,000 Negotiable
- Location: Japan: 7F, Tohshin Aoyama Building, 2-10-13 Shibuya, Shibuya-ku, Tokyo, Japan; US: 651 N. BROAD ST., SUITE 201, MIDDLETOWN DE 19709
- Employment Conditions:
- Full-time ()
- Original working hours
- Holidays:
- Annual Holidays: 2 days off per week (Saturdays and Sundays), national holidays, year-end and New Year holidays
- Paid Annual Leave: 10 or more days per year (varies depending on the month of joining)
- Probation Period: 3 months
- Benefits:
- Health Insurance
- Welfare (Pension) Insurance
- Employment Insurance
- Workers Compensation Insurance
- Commute allowance: up to ¥30,000 per month
- Stock Option Rewards System: According to company regulations
- Visa Support