Salil Borkar

to give engineers real-time SQL cost estimation via offline AST parsing — runaway cloud spend caught before deployment, with zero latency and zero PII risk.

The execution worked. The distribution model didn't. Most enterprises block unapproved VS Code extensions, so the engineers who needed cost guardrails most couldn't install them. That insight drove the pivot to llm-cost-lint.

pull requests for AWS Bedrock and Azure OpenAI API calls, estimates per-call and projected monthly cost, and posts the breakdown as a PR comment before code merges. Same shift-left FinOps thesis as Cost Sentinel — moved from the IDE (where enterprises block it) to CI/CD (where they actively want it).

Open source. MIT licensed. Available on the GitHub Actions Marketplace.

— market sentiment, price analysis, and SEC 10-K filing intelligence — orchestrated by a Gemini-powered reasoning loop. The orchestrator dynamically sets each agent's inputs based on prior agent outputs: sentiment themes drive the chart date range and generate targeted RAG questions.

Shared context bus, structured audit trail, and Streamlit UI. Fully local embeddings (sentence-transformers + ChromaDB). 10-K filings auto-downloaded from SEC EDGAR.