Skysight, Inc.
Skysight, Inc. is an AI infrastructure company building tools that increase human leverage by making large-scale generative-model data processing more efficient. Its Sutro platform helps teams turn expert decisions into reliable AI Functions and run them at scale.
At a Glance
- Applied AI teams
- Data and ML engineering teams
- AI training-data teams
- Companies using LLMs for large-scale data processing
- +2 more
AI Tools by Skysight, Inc.
(1)
Sutro
AI Function Optimization Platform
Discussions
No discussions yet
Be the first to start a discussion about Skysight, Inc.
Latest News
Sutro updated its Terms of Service, formally identifying the provider as Skysight Inc. doing business as Sutro.
The public Sutro SDK repository recorded a foundation-sync commit and continued active development.
Sutro Python SDK version 0.1.60 was released on PyPI.
Sutro Python SDK version 0.1.59 was released on PyPI.
Products & Services
Task-specific judges, classifiers, routers, matchers, and structured extractors aligned to customer decision preferences. Users upload representative data, review difficult cases, add rationales, optimize prompts, compare deployment models, and publish reusable Functions for online or batch execution.
Serverless asynchronous inference for high-volume workloads, supporting Sutro Functions and pre-trained open-source LLMs with usage-based pricing, DataFrame-friendly inputs and outputs, observability, estimates, and downloadable results.
Apache-2.0 Python client and CLI for running Functions, batch inference, embeddings, multi-model comparisons, job management, quotas, and result downloads.
A public collection of synthetic-human datasets published on Hugging Face in 10k, 50k, and 1m variants.
Market Position
Sutro positions itself between prompt engineering/fine-tuning and conventional evaluation or monitoring products: it uses expert feedback to directly shape a repeatable task-specific Function, measures it against held-out and difficult cases, and then runs it in batch or as an event stream. It emphasizes lower-cost, high-throughput, reproducible inference and claims differentiation from evals products that only monitor quality. Relevant alternatives include general-purpose LLM APIs and batch providers, model-serving/inference platforms, prompt-optimization tools, and LLM evaluation/observability products.
Leadership
Founders
Seth Kimmel
Founder/CEO of Sutro; an engineer and technologist who has worked in product and engineering at Shaper Capital, provided data/ML/AI consulting, and worked as a software/data engineer. His public profile identifies him as based in San Francisco and building Sutro since August 2024.
Executive Team
Seth Kimmel
Founder and CEO
Engineer and technologist; previously worked in product and engineering at Shaper Capital, in data/ML/AI consulting, and as a software/data engineer.
Founding Story
Skysight describes its initial vision as enabling the productivity, efficiency, and discovery gains of generative models by making data processing work at very large scale practical. The company began with large-scale AI-inference infrastructure and later presented Sutro as a product for encoding an organization's expert judgment into repeatable AI Functions.
Business Model
Revenue Model
Sutro combines platform subscriptions with inference/compute usage. The public site offers platform access, while Batch uses machine-time pricing so customers pay for the compute their workloads consume; enterprise, self-hosted, and bespoke data-processing arrangements are tailored to workload and deployment needs.
Pricing Tiers
Includes $100/month in inference credits.
Tailored to customer needs and scale.
The site illustrates approximately $0.009 per 1K records for batch and $0.012 per 1K records for single execution for an example Function; actual cost depends on workload.
Target Markets
- Applied AI teams
- Data and ML engineering teams
- AI training-data teams
- Companies using LLMs for large-scale data processing
- Organizations with strict data privacy and security requirements
- Enterprise teams requiring self-hosted or isolated-cloud deployments
- Agent and model evaluation, QA gates, and rubric-based judging
- Lead qualification, ticket routing, ownership, prioritization, and escalation
- Document categorization and structured extraction
- Company, person, product, and incident matching and resolution
- Trust, safety, fraud, spam, compliance, KYC, risk, and eligibility decisions
- Dataset filtering, labeling, tagging, enrichment, and data quality normalization