Architecture decisions from 0 to 1,
scaling from 1 to N
A lean team with a short decision chain, where new members directly shape core technology and business decisions.
No meaningless meetings, just real challenges and high-density growth.
What We Offer
This is not an ordinary job; it's a chance to help define the next form of AI infrastructure.
Full Tech Stack
From the compute base to Agent applications, the density of real technical challenges is high. You won't maintain legacy systems here; you'll help build them.
Flat Organization
Short decision chains and real say over architecture for engineers. Your judgment directly shapes system design, without waiting on layers of approvals.
Ample Resources
Heterogeneous compute (NVIDIA / AMD / domestic GPU / NPU) and real production traffic, so engineering problems are not toy scenarios.
Competitive
Competitive salary and equity incentives, with growth opportunities in both Hong Kong and mainland China. Early core members share in the company's growth.
Open-Source Friendly
We support open-source contributions and public technical sharing. The work you do here can become a portfolio piece you showcase publicly.
Open Positions
Click a role card for details, or send your resume to lionelLiu@aceproinv.com
Background
The platform already has a stable model supply and price advantage; it needs dedicated sales to convert capacity into sustained commercial consumption. Target customers: AI-native applications and Agent vendors, SaaS / software companies, mid-to-large internet companies, going-global companies, and system integrators.
Responsibilities
- 01Own the full B2B sales cycle for mainstream LLM Tokens
- 02Build the customer map and sales funnel
- 03Deeply understand customer scenarios (chat, RAG, Agent, batch, multimodal, etc.)
- 04Design pricing and commercial terms
- 05Maintain channel and ecosystem partnerships
- 06Continuously feed back market dynamics to shape product and pricing strategy
Requirements
- —Bachelor's degree or above, 3+ years of B2B sales with cloud / API / compute product sales cases
- —Able to hold substantive technical conversations with CTOs, architects, and algorithm leads
- —Strong ability to independently develop new customers, results-driven and resilient
Nice to Have
Background
The distribution platform is a core company asset, benchmarked against OpenRouter / LiteLLM-class systems. We need 3 engineers across three directions: gateway & routing, metering & billing, and console & developer experience.
Responsibilities
- 01Design and implement a high-concurrency, low-latency model API gateway
- 02Multi-provider aggregation and intelligent routing
- 03Build the metering, quota, and billing system
- 04Developer console and API key management
- 05Observability system and risk-control policies
- 06Capacity planning and full-chain load testing
Requirements
- —Bachelor's degree or above in CS, 3+ years of backend development
- —Proficient in at least one of Go / Rust / Python
- —Familiar with gateway fundamentals: HTTP/2, SSE, rate limiting, caching
- —Familiar with K8s / Redis / Kafka / PostgreSQL / ClickHouse
- —Understands Token billing models and API product design
Nice to Have
Background
Per-token cost directly determines margin. The role spans three levels, inference engine, operators, and cluster, with the goal of continuously lowering per-token cost, one of the company's core competitive advantages.
Responsibilities
- 01Production deployment and tuning of mainstream open-source models (Llama / Qwen / DeepSeek, etc.)
- 02Inference platform optimization: vLLM / SGLang / TRT-LLM, PD separation, KV Cache, speculative decoding, quantization
- 03Operator optimization: CUDA / Triton / CUTLASS performance tuning
- 04Cluster optimization: TP / PP / EP / DP parallelism, NCCL / RDMA communication optimization
- 05Build load-testing and evaluation systems, continuously tracking cost and SLO
- 06Domestic GPU / NPU adaptation (Ascend / Cambricon / Moore Threads, etc.)
Requirements
- —Bachelor's or above in CS / microelectronics / math, 3+ years in HPC or deep-learning systems
- —Proficient in Python + C++, with CUDA programming ability
- —Deep understanding of Transformer inference (attention, KV Cache, quantization)
- —Hands-on use or development experience with vLLM / SGLang / TRT-LLM
- —System performance analysis and profiling ability
Nice to Have
Background
Agents are the main growth engine for Token consumption. We are building the runtime and orchestration for multi-Agent collaboration, the platform-layer engineering direction closest to business value.
Responsibilities
- 01Agent Runtime development: context management, memory systems, sandbox isolation, lifecycle control
- 02Model-tool collaboration: MCP / tool calling / RAG / model routing
- 03Task orchestration: multi-Agent / DAG workflows, resume, human-in-the-loop
- 04Observability & evaluation: trace, span, Agent behavior assessment
- 05End-to-end joint optimization with the inference platform to lower Agent-scenario Token cost
- 06Explore and implement Agent interconnection protocols (A2A and other emerging standards)
Requirements
- —Bachelor's or above in CS, 3+ years in backend or platform development
- —Proficient in Python or Go with solid distributed-systems engineering
- —Hands-on Agent / LLM application experience, understanding prompt engineering and context management
- —Familiar with mainstream Agent frameworks (LangGraph / AutoGen / CrewAI / Temporal, etc.)
Nice to Have
Don't see a fitting role?
If you believe you can add value to the team, feel free to reach out. We're always open to great people.