Third Intelligence Tech Stack

AI research lab building next-generation intelligence systems and AGI

Software Development Chiyoda-ku, Tokyo 11–50 employees Founded 2025 Privately Held

Third Intelligence is a Tokyo-based AI research company founded in 2025, operating a Python + PyTorch + CUDA stack with infrastructure on AWS and GCP. The hiring mix—five research roles, one design, one engineering—and project portfolio (SFT/RLHF fine-tuning, synthetic data pipelines, native voice models, multimodal understanding) reveal a research-first organization actively bridging the research-to-product gap. Pain points around model alignment, low-latency voice interaction, and scaling inference suggest they're moving beyond algorithmic research toward deployable systems.

Tech Stack 15 technologies

Core StackPython PyTorch C++ Go AWS LangChain Docker Kubernetes Figma DeepSpeed CUDA ASIC GCP OpenAI API LlamaIndex

What Third Intelligence Is Building

◆Challenges

Research to product pipeline
Balancing research and product integration
Model alignment and safety
Low-latency voice interaction
Implementing research models into production
Building scalable ai application environment
Scaling next-gen ai models
Optimizing custom inference engine
Strengthening research organization
Translating research into product value

▲Active Projects

Sft and rlhf fine-tuning
Synthetic data pipeline
Custom evaluation metrics
Next-generation ai algorithm research
Low-latency end-to-end voice dialogue pipeline
Multimodal context understanding system
Native voice model research
Integrating llms into product
High availability low latency backend architecture
Data pipeline for model inference

Hiring Activity

Accelerating7 roles · 7 in 30d

Department

Research

Design

Engineering

Seniority

Mid

Senior

Director

Company intelligence

Find more companies like Third Intelligence by tech stack, pain points and active projects

Get started free

About Third Intelligence

Third Intelligence develops AI systems and algorithms aimed at advancing toward AGI. The company is based in Tokyo with 11–50 employees, structured primarily around research and engineering teams. Their active work spans model fine-tuning, synthetic data generation, custom evaluation metrics, voice dialogue systems, and multimodal reasoning—all underpinned by open-source frameworks (PyTorch, LlamaIndex, LangChain) and cloud infrastructure. Early-stage hiring velocity is accelerating, with seven roles open across research, engineering, and design, weighted toward mid-to-senior technical staff.

HeadquartersChiyoda-ku, Tokyo

Company Size11–50 employees

Founded2025

Hiring MarketsJapan

Frequently Asked Questions

What is Third Intelligence's tech stack?

Python, PyTorch, DeepSpeed, CUDA, C++, Go, AWS, GCP, OpenAI API, LangChain, LlamaIndex, Docker, Kubernetes, and Figma. ASIC development is also in use.

What is Third Intelligence working on?

SFT and RLHF fine-tuning, synthetic data pipelines, custom evaluation metrics, next-generation AI algorithm research, low-latency voice dialogue systems, multimodal context understanding, native voice models, and production-grade backend architecture.