What we are not
- Not a middleman reselling model APIs
- Not slide-deck consultants who hand over a PPT and walk away
- Not "AI researchers" chasing the latest hype
- Not a "solutions vendor" that lives off big-name endorsements
ABOUT · Who we are
We are Banfang,
We are Banfang,
half belief, half engineering.
An AI veteran founded in Hangzhou in 2018. The name "Banfang" means "one half": one half is belief — we believe AI will reshape most industries — and the other half is engineering — the rigorous discipline that turns that imagination into reality.
What we are
- An engineering team that actually writes code and stands behind it in production
- A partner who will debug alongside you and watch the business metrics with you
- Almost obsessive about the word "delivery"
- Willing to act as your in-house "AI CTO"
01
Outcomes first
We do not chase demos that look good but do nothing. We care about systems that run in production and earn a renewal.
02
Engineering first
Evaluation sets, monitoring, rollback, graceful degradation, auditing — we make these the foundation of every AI project.
03
Long-term
Every project we take on should become a capability that stays inside the client, not a one-off transaction.
TIMELINE · Key milestones
How we got here
2018 · Founded
Banfang Technology is founded in Hangzhou
The founding team came from leading internet firms and AI labs, starting out in NLP, conversational AI and search ranking. We were among the earliest teams exploring "enterprise conversational intelligence."
2019 – 2021 · Deep in NLP
Delivered conversation and knowledge systems for 20+ companies
Over these years we pushed semantic search, intent recognition and FAQ bots as far as they could go. We also felt, firsthand, the ceiling of "traditional NLP" in real business, and began studying large language models in earnest.
2022 · All in on LLMs
A full pivot to LLMs and Agents
The month GPT-3.5 came out, we shut down a still-profitable legacy NLP product line and moved our entire engineering core onto large-model Agents. That year, we made the right call.
2023 Q4 · Agent Duoduo v1
Our first support Agent "with memory" goes live
We got a long-term-memory support system running on-site at a retail client, and the human handoff rate dropped 58% in the first week.
2024 · Moss is born
Moss, the autonomous AI employee, launches
It started with one enterprise client asking "can you write our competitor report automatically every week?" From there, Moss grew into an autonomous Agent that handles hundreds of real-world tasks.
2025 · Scaling up
Past 40 enterprise clients
Our work now spans six industries: retail, finance, manufacturing, legal, education and SaaS. Median annual revenue per client crossed the seven-figure mark.
2026 · In progress
We are building something bigger
We are combining Moss's autonomy, Hermes's memory and skill accumulation, and the LLM Wiki paradigm proposed by Karpathy into an "AI employee operating system" that every company can afford. Come join us, or become our 100th client.
Partner with us, invest, or join us
Email or call us anytime. We would love to meet fellow travelers.
📧 bd@thebanfang.com 📞 +86 187 0117 8691