U
Upstage operates a dual-engine AI platform: Solar, a cost-optimized LLM, and Document AI, a specialized OCR and document-processing system. The tech stack reveals a RAG-first approach (LangChain, LangGraph, Hugging Face, PyTorch) with active infrastructure scaling (Kubernetes, multi-region architecture design). Product and research teams (16 headcount combined) drive model iteration, while engineering velocity is decelerating—likely reflecting a shift from hiring breadth to deepening document automation and LLM fine-tuning.
Tech Stack 55 technologies
Core StackLangChain Hugging Face Python AWS Salesforce HubSpot Snowflake Docker Kubernetes Prometheus Grafana Java Spring JavaScript TypeScript PyTorch FastAPI React Next.js Storybook LangGraph OpenCV NLTK spaCy Excel Google Sheets GCP Azure Tomcat Git+25 more
What Upstage Is Building
◆Challenges
- Reducing document processing time
- Automating data pipelines
- Scaling up education solutions
- Optimizing model performance
- Automating document processing
- Optimizing evaluation systems
- Document workflow problems
- Existing benchmark limitations
- Llm post-training performance
- Partner marketing collaboration
▲Active Projects
- Ai model data pipeline development
- Information extract api
- Dataset augmentation and human feedback loop
- Global multi-region infrastructure architecture design
- Document ai product marketing
- Private llm service sales
- Prototype demos and pocs for real use cases
- Poc prototype design and development
- Cloud solution consulting for global customers
- Partner system integration and performance optimization
Hiring Activity
45 roles · 10 in 30d
Department
Seniority
Company intelligence
