Senior NLP Engineer

About the job

Position Summary We're seeking a Senior NLP Engineer with a hacker mindset to revolutionize our information extraction engine for financial/legal data processing. This is a product-driven role focused on delivering production-ready solutions where accuracy and speed of output are paramount. Our systems are customer-facing with Human-in-the-Loop (HITL) workflows, requiring optimization for seamless human-AI collaboration. You'll have complete ownership from problem analysis to production deployment, leveraging any LLM, technique, or creative approach that delivers maximum accuracy in minimum time while ensuring optimal user experience. Key Responsibilities Core Product-Driven AI Work • Enhance Information Extraction Engine: Redesign and optimize our current system using state-of-the-art LLMs for financial/legal document processing with focus on production accuracy and speed. • Accuracy Optimization: Achieve highest possible extraction accuracy through any means necessary—fine-tuning, prompt engineering, ensemble methods, or hybrid approaches. • LLM Integration: Implement and experiment with various LLMs (GPT, Claude, Llama, Gemini, etc.) to find optimal solutions for production use. • Creative Problem Solving: Think like a hacker —if traditional ML doesn't work, try prompt engineering, RAG, few-shot learning, or completely novel approaches. • LLM Development: Lead the pre-training and fine-tuning of large language models (LLMs) to optimize performance for specific financial/legal use cases. • Information Retrieval, Extraction & Classification: Develop and implement techniques for retrieving, extracting, classifying, and ranking large-scale financial/legal datasets using advanced algorithms and vector databases. • Production-First Mindset: Every solution must be production-ready with measurable accuracy and performance metrics. • Time-to-Accuracy Optimization: Balance between achieving high accuracy and delivering results within acceptable time constraints. Human-in-the-Loop (HITL) Workflow Optimization • HITL System Design: Build AI systems optimized for human review, correction, and validation workflows. • Confidence Scoring & Routing: Develop intelligent routing systems that send low-confidence extractions to human reviewers while auto-approving high-confidence results. • Interactive Correction Interfaces: Design systems that learn from human corrections and feedback in real-time. • Workflow Efficiency: Optimize human review processes to minimize time spent while maximizing accuracy improvements. • Active Learning Integration: Implement systems that strategically request human input on the most valuable data points. • Feedback Loop Optimization: Build mechanisms to continuously improve AI performance based on human corrections and preferences. • User Experience Design: Ensure seamless handoffs between AI processing and human review stages. Engineering Responsibilities • Third-Party LLM API Integration: Seamlessly integrate and manage multiple LLM APIs (OpenAI, Anthropic, Google, etc.) with fallback mechanisms. • API Development & Management: Build robust APIs for information extraction services with proper error handling and monitoring. • System Evaluation & Benchmarking: Develop comprehensive evaluation frameworks to measure accuracy, latency, and cost across different LLM approaches. • Performance Engineering: Optimize systems for low-latency, high-throughput information extraction. • Integration Architecture.

Requirements

• NLP
• Information Extraction
• LLM
• API Integration
• Machine Learning

Preferred Technologies

• NLP
• Information Extraction
• LLM
• API Integration
• Machine Learning