

NLP Engineer
⭐ - Featured Role | Apply direct with Data Freelance Hub
This role is for a Founding NLP Engineer with over 3 years of experience in transformer-based architectures, proficient in PyTorch and token classification. It's a remote position with equity-based compensation, focusing on building modular AI components for a patent-pending system.
🌎 - Country
United States
💱 - Currency
$ USD
-
💰 - Day rate
-
🗓️ - Date discovered
August 19, 2025
🕒 - Project duration
More than 6 months
-
🏝️ - Location type
Remote
-
📄 - Contract type
Unknown
-
🔒 - Security clearance
Unknown
-
📍 - Location detailed
United States
-
🧠 - Skills detailed
#Classification #GitHub #Debugging #AI (Artificial Intelligence) #IP (Internet Protocol) #PyTorch #"ETL (Extract #Transform #Load)" #NLP (Natural Language Processing) #Datasets
Role description
Founding NLP Engineer – Confidential AI Architecture
Company: Confidential | Patent-Pending System
Location: Remote-first (SF / NY / LA optional)
Type: Full-Time | Founding Role
Compensation: Equity-based at this stage – substantial long-term upside
Contact: Founder@whitebeltstudio.com
About the Role:
We’re building a proprietary AI system from first principles, modular, interpretable, and deeply original.
This is not prompt engineering. Not model-wrapping. You’ll be working on core AI behavior, designing structured components that operate precisely and independently inside a novel architecture. Every part you build will become IP.
You’ll collaborate directly with the founder to shape a system focused on clarity, performance, and generalization, not convenience.
Your Responsibilities:
• Train and refine transformer-based model components (e.g., RoBERTa, DeBERTa)
• Design high-accuracy token-level classification heads for structured tasks
• Build out custom evaluation pipelines and modular feedback loops
• Work with both internal and public datasets (e.g., UD, OntoNotes)
• Optimize data strategies: rare label handling, sampling, augmentation
• Build components that integrate into a larger structured system
Ideal Candidate Profile:
You’re an AI builder, not just a researcher. You know how to get under the hood of a transformer. You understand sequence labeling and want to build clean, accurate components, not throw everything at a monolith and hope.
You likely bring:
• 3+ years of experience with transformer-based architectures
• Expertise in PyTorch, HuggingFace, and token classification tasks
• Experience debugging tokenization errors, class imbalance, label noise
• Preference for modular systems over end-to-end pipelines
• Ability to build models that explain themselves not just perform
Bonus experience:
• Working with multi-head or cascaded model designs
• Benchmarking against structured NLP datasets
• Prior work involving emotion, tone, or symbolic+neural systems
• Experience at a 0→1 AI project or research-first startup
Why Join:
• Be part of a confidential, IP-protected project at the ground level
• Work directly with the founder on core system design
• Influence not just model behavior—but the entire system paradigm
• Build something meant to last: not a feature sprint, but a foundation
• Significant equity ownership in a patent-pending, high-leverage system
How To Apply:
Send a short note about your background and interest, plus any relevant work (GitHub, papers, projects) to:
Founder@whitebeltstudio.com
Founding NLP Engineer – Confidential AI Architecture
Company: Confidential | Patent-Pending System
Location: Remote-first (SF / NY / LA optional)
Type: Full-Time | Founding Role
Compensation: Equity-based at this stage – substantial long-term upside
Contact: Founder@whitebeltstudio.com
About the Role:
We’re building a proprietary AI system from first principles, modular, interpretable, and deeply original.
This is not prompt engineering. Not model-wrapping. You’ll be working on core AI behavior, designing structured components that operate precisely and independently inside a novel architecture. Every part you build will become IP.
You’ll collaborate directly with the founder to shape a system focused on clarity, performance, and generalization, not convenience.
Your Responsibilities:
• Train and refine transformer-based model components (e.g., RoBERTa, DeBERTa)
• Design high-accuracy token-level classification heads for structured tasks
• Build out custom evaluation pipelines and modular feedback loops
• Work with both internal and public datasets (e.g., UD, OntoNotes)
• Optimize data strategies: rare label handling, sampling, augmentation
• Build components that integrate into a larger structured system
Ideal Candidate Profile:
You’re an AI builder, not just a researcher. You know how to get under the hood of a transformer. You understand sequence labeling and want to build clean, accurate components, not throw everything at a monolith and hope.
You likely bring:
• 3+ years of experience with transformer-based architectures
• Expertise in PyTorch, HuggingFace, and token classification tasks
• Experience debugging tokenization errors, class imbalance, label noise
• Preference for modular systems over end-to-end pipelines
• Ability to build models that explain themselves not just perform
Bonus experience:
• Working with multi-head or cascaded model designs
• Benchmarking against structured NLP datasets
• Prior work involving emotion, tone, or symbolic+neural systems
• Experience at a 0→1 AI project or research-first startup
Why Join:
• Be part of a confidential, IP-protected project at the ground level
• Work directly with the founder on core system design
• Influence not just model behavior—but the entire system paradigm
• Build something meant to last: not a feature sprint, but a foundation
• Significant equity ownership in a patent-pending, high-leverage system
How To Apply:
Send a short note about your background and interest, plus any relevant work (GitHub, papers, projects) to:
Founder@whitebeltstudio.com