Stochastic’s cover photo
Stochastic

Stochastic

Software Development

New York, NY 1,169 followers

Autonomous AI Built for the Enterprise of Tomorrow

About us

Stochastic is building a full-stack reasoning agent platform to automate administrative workflows in regulated industries, starting with healthcare.

Website
https://www.stochastic.ai
Industry
Software Development
Company size
2-10 employees
Headquarters
New York, NY
Type
Privately Held

Locations

Employees at Stochastic

Updates

  • Stochastic reposted this

    View profile for Glenn Ko

    Stochastic4K followers

    This is an insightful analysis by Jaya Gupta on why capturing the context behind AI decisions is critical and why current infrastructure like Databricks and Snowflake still falls short. For AI agents in regulated industries, these capabilities are table stakes. Beyond a simple context graph, companies require robust audit trails that capture the specific versions of inputs, policies, and exceptions used, alongside a clear record of any human intervention. At Stochastic, we ship our products fully equipped with context graphs, versioning, and recorded history of every manual intervention. If you'd like to learn more, shoot me a message or leave comments below.

    View profile for Jaya Gupta

    Partner @ Foundation Capital

    There's a subtle infrastructure problem no one's talking about. Data warehouses like Snowflake and Databricks are in the read path. They receive data via ETL after decisions are made. By the time it lands, the decision context is gone. They can tell you what happened, but they can't tell you why. For example let's say a support lead checks customer ARR in Salesforce, sees two open escalations in Zendesk, reads a Slack thread flagging churn risk, and decides to escalate. The synthesis of that process happens in their head. The ticket just says "escalated to Tier 3." The warehouse gets "escalated to Tier 3" three hours later, but the reasoning doesn't really get recorded. This matters for agents. If you want an agent to make that same escalation decision, it doesn't just need access to the data. It needs access to how that data was synthesized into a decision the last ten times. The precedents. The exceptions. The judgment calls. That requires being in the write path, at commit time and in the execution loop. That's the only place you can capture the full context: what inputs were gathered, what policy applied, what exception was granted, who approved. If you persist those traces, you get something we call a context graph (not embeddings), but more a structured record of how decisions were actually made, stitched across entities and time, so precedent becomes queryable. The startups sitting in the agent orchestration layer are the ones positioned to build this. Not because they're smarter. Because they're in the right place architecturally. Read more about our thesis below and if you are building a company in this space, excited to chat!

  • Stochastic reposted this

    View profile for Glenn Ko

    Stochastic4K followers

    Another impressive result from DeepSeek AI. Their latest R1 update (R1-0528) is now the best open source reasoning model, second only to OpenAI-03 on most benchmarks. What does this mean for the enterprises adopting AI? While the open source models are trailing the best closed source models, the gap is closing (already better than most closed models) and soon the models themselves may not be the differentiator anymore. We already see signs of foundation model providers striving to move into application space or trying to close more enterprise deals before the doors are shut. If you can choose any models of similar performance, why would anyone choose closed models over open ones? With close models, you have better 1. Control - no outages of APIs beyond your control that freezes your application 2. Governance - host it yourself and everything will be under your umbrella 3. Cost - as you scale, you don't have to keep paying per tokens. If you buy your own GPUs, its even cheaper 4. Latency - no transmission latency between APIs 5. Fine-tuning - as the models continue to improve, fine-tuning is often overlooked but ask around. All serious enterprises are doing some fine-tuning already on some of the models in production. The question is when should the enterprises pull the trigger?

    • No alternative text description for this image

Similar pages

Browse jobs

Funding

Stochastic 4 total rounds

Last Round

Pre seed
See more info on crunchbase