Alphaset
Show HN: Would a learning-first AI evaluation platform be useful?
Last verified:
What is Alphaset?
Alphaset is an AI data company that builds high-quality post-training data for frontier AI models and enterprises. They specialize in creating organic, expert-grown datasets that are essential for fine-tuning more capable and reliable AI systems. Their engineers work directly with AI teams to identify where models are breaking and deliver batches of data that pass the client's evaluation before delivery.
The company offers three core services: finding domain experts for areas clients can't staff, building failure cases against evaluations that aren't moving, and creating custom annotation tools tailored to specific needs. Every batch of data is run against the client's eval before delivery to ensure quality. The experts working on projects are directly employed by Alphaset, not subcontracted.
Alphaset is designed for AI research labs, enterprise AI teams, and startups building frontier language models that need high-quality human-annotated data for post-training techniques like supervised fine-tuning and reinforcement learning. The company is YC W25-backed and was founded by former xAI iOS engineer Yuhao Chen who contributed to Grok Imagine and Grok 4.1, along with co-founder Tielong Tangeleih who has experience building AI products across gaming and health apps.
Alphaset pricing
Pricing model: Freemium
Pricing details are not publicly listed on the website. The site states 'If you're looking for a high quality dataset, contact us' and features a 'Book a demo' button, indicating custom enterprise pricing. Clients need to contact the company directly to discuss pricing and project scope.
Alphaset pros
- Builds high-quality post-training data for frontier AI models
- Engineers sit with your team to identify model failure points
- Every batch run against client eval before delivery
- Experts are directly employed, not subcontracted
- Finds domain experts for hard-to-staff areas
- Builds custom annotation tools on demand
- Specializes in building failure cases against evals
- YC W25-backed company with strong founder background
- Founder previously worked at xAI on Grok Imagine
- Scales from simple datasets to complex domain-specific data
- Focuses on organic, human-created data not synthetic
- Works across multiple domains requiring expertise
- Direct collaboration with client AI engineering teams
- Quality guarantee through eval testing before delivery
- Experienced team with AI training and product background
Alphaset cons
- No public pricing information available
- Contact required to get pricing or start a project
- No free tier mentioned on website
- Appears to be enterprise-focused, not self-service
- No public case studies or customer testimonials
- Limited information about turnaround time
- No mention of data formats or API access
- Company appears to be early-stage (YC W25)
Frequently asked questions about Alphaset
What type of data does Alphaset build?
Alphaset builds post-training data for AI models, including high-quality human-annotated datasets for supervised fine-tuning and reinforcement learning. They create domain-specific expertise data covering areas like finance, software engineering, medicine, and legal, as well as nuanced reasoning data and vertical computer use data.
How does Alphaset ensure data quality?
Every batch of data is run against the client's evaluation before delivery. Their engineers sit with the client's team to figure out where the model is breaking, ensuring the data directly addresses model weaknesses. This quality control process guarantees the data moves the eval.
Are the experts subcontracted?
No, the experts are yours and not subcontracted. Alphaset directly employs the experts working on your project, ensuring better quality control and consistency.
What domains can Alphaset provide experts for?
Alphaset can find experts for any domain you can't staff for, including specialized fields like finance, software engineering, medicine, legal, and other areas requiring domain-specific expertise. They specialize in hard-to-staff domains.
Can Alphaset build custom annotation tools?
Yes, Alphaset builds custom annotation tools tailored to your specific needs. If you need a custom annotation tool for your post-training data pipeline, they will build it for you.
What is Alphaset's approach to synthetic data?
Alphaset focuses on organic, expert-grown human data rather than synthetic data. They position themselves as an anti-synthetic data platform, recognizing that high-quality human data is essential for creating near-perfect AI systems.
Who is Alphaset's target customer?
Alphaset serves foundational model labs, enterprises, and startups that need high-quality human-generated datasets for fine-tuning AI systems. Their customers include AI research teams building frontier language models that require domain-specific expertise and nuanced reasoning data.
What is Alphaset's background?
Alphaset is a YC W25 company founded by Yuhao Chen (former xAI iOS engineer who contributed to Grok Imagine and Grok 4.1, core agent engineer at Manus AI) and Tielong Tangeleih (built across AI, gaming, and enterprise BD, worked on B2C AI health app).
How do I start working with Alphaset?
You need to contact Alphaset directly through their website. They have a 'Book a demo' button and state 'If you're looking for a high quality dataset, contact us' to initiate discussions about your data needs.
What if my eval isn't improving?
Alphaset specializes in building failure cases against evals that aren't moving. They will analyze where your model is breaking and create targeted data batches specifically designed to move your eval forward.