Machine Learning Tech Visualizer, Designer & Writer
Etched
About Etched
Etched is building AI chips that are hard-coded for individual model architectures. Our first product (Sohu) only supports transformers, but has an order of magnitude more throughput and lower latency than a B200. With Etched ASICs, you can build products that would be impossible with GPUs, like real-time video generation models and extremely deep chain-of-thought reasoning.
Machine Learning Researcher - Reasoning & Understanding
With Etched ASICs, we have fundamentally different constraints than existing AI chips. We don’t have the same batch-size-latency tradeoff as GPUs. We can parallelize workloads and digest large contexts much more efficiently than GPUs.
Sohu enables entirely new research directions and products. When our chips come out, these use cases need to already be mature and visceral. Whether real-time video, agents, speculative decoding, or new tree search algorithms, we must create the market for our hardware.
In this role, you will lead efforts on research directions that trade compute for improved understanding and speed, particularly with agents and new search techniques.
Representative projects:
- Combine multi-agent with RAG to improve the quality of QA
- Evaluate new system on standard benchmark
- Design new and verifiable benchmark for agent reasoning
- Design LLM content understanding based recommendation systems
You may be a good fit if you:
- Past projects or publications with substantial impact in ML and/or CV (quality > quantity)
- Strong hands-on engineering skills, particularly with python, pytorch, CUDA, DDP/FSDP
- Deep understanding of open and closed source model architectures and open source libraries for transformer training and inference
- Familiarity with LlamaIndex, LangChain, MetaGPT, CoT, ToT, and beam search
- Ability to think outside the box and make tradeoffs considering feasibility, quality, and time-to-ship of a project
We encourage you to apply even if you do not believe you meet every single qualification.
Strong candidates may also have experience with:
- Knowledge graphs
- Name entity recognition
- Tool calling and coder LLMs
- SWE bench & SWE agent
How we’re different:
Etched believes in the Bitter Lesson. We think most of the progress in the AI field has come from using more FLOPs to train and run models, and the best way to get more FLOPs is to build model-specific hardware. Larger and larger training runs encourage companies to consolidate around fewer model architectures, which creates a market for single-model ASICs.
We are a fully in-person team in Cupertino, and greatly value engineering skills. We do not have boundaries between engineering and research, and we expect all of our technical staff to contribute to both as needed.
Benefits:
- Full medical, dental, and vision packages, with 100% of premium covered, 90% for dependents
- Housing subsidy of $2,000/month for those living within walking distance of the office
- Daily lunch and dinner in our office
- Relocation support for those moving to Cupertino