AI & Machine Learning · sub-niche
LLM observability stacks.
Production AI apps need traces, evals, and cost dashboards — the Datadog of AI is still being decided.
Why now
Every product team is shipping an AI feature. Most are blind to cost, latency, and quality regressions. The observability layer is being chosen now and switching costs are high.
What the signal looks like
Repos with OpenTelemetry instrumentation, SDK adapters for the top five model providers, and a hosted dashboard demo in the README.
Public examples
We name publicprojects + categories only — never founders we track inside the paid product. The buyer’s edge stays inside the product.
- Langfuse-style hosted observability platforms
- Helicone-shaped lightweight proxies
- Phoenix from Arize — open-source observability for LLMs
What this displaces
Console.log statements and weekly spreadsheet exports.
Our build-vs-invest call
The product is hosted, not open-source-only. Win the developer relationship with a 10-line SDK install, monetize on retention + alerts. Compete on the eval primitives, not the trace viewer.
Common questions about this niche
- Is Datadog going to win this?
- Eventually they'll be a credible option. The 18-month window belongs to focused teams that own the workflow.
- What's the funding signal?
- Free-tier-to-paid conversion above 4%. That's the metric every Tier-1 firm asks about.
- How does this compound?
- Traces → evals → model routing → cost optimization. Each adjacency is a new product line.
More inside AI & Machine Learning
- LLM eval harnesses — Reproducible eval suites that an AI-native team can drop into CI and trust by lunchtime.
- Agent orchestration frameworks — The 'LangChain for X' slot is still wide open — pick a vertical, ship the runtime, win the wedge.
- Retrieval-augmented search libraries — RAG-as-a-library — bring-your-own embedding, bring-your-own vector store, win on developer ergonomics.
- Fine-tuning tools for non-ML teams — Take fine-tuning out of the notebook. Product teams want to point at JSONL and get a deployable adapter.