Tech Stack
Tools, frameworks, and data engines used across production deployments and applied research projects — categorized by architectural role, with references to implementation and evaluation scope.
Production AI, Agents & Applied MLOps
Agent workflows need explicit permissions, observability and evaluation. The inventory includes production tools and research candidates; an entry does not imply that every listed version has passed production validation. Project status lives in Active Labs, and current adoption conditions in Emerging Tech.
| Role | Tools | Use / evaluation scope |
|---|---|---|
| Agentic frameworks & runtimes | LangGraph, Pydantic AI, Google ADK (2.0 workflow model), OpenAI Agents SDK, Claude Agent SDK, LlamaIndex, FastAPI | Stateful multi-agent supervision in Agent Fabric |
| Agent protocols & model access | Model Context Protocol (MCP, 2026-07-28 specification), A2A v1.0, Vertex AI Agent Engine, Amazon Bedrock AgentCore, Gemini, Anthropic Claude, Hugging Face, LiteLLM | Remote tools, cross-runtime delegation, and cost-aware inference routing in Agent Fabric and Inference Fabric |
| Inference serving & routing | SGLang, vLLM, TensorRT-LLM, NVIDIA Dynamo (disaggregated prefill/decode over NIXL), LMCache, llama.cpp, Ollama | Local-first serving, EAGLE-3 speculative decoding, and tiered KV cache in Inference Fabric |
| Open-weight models & guardrails | gpt-oss, Qwen, Gemma, DeepSeek, Llama, Hermes; Llama Guard, gpt-oss-safeguard, Qwen3Guard | Utility-tier routing in Inference Fabric; policy-following safety classification in Agent Fabric |
| Vector, graph & memory storage | Qdrant, pgvector, LanceDB (Lance), Milvus, LadybugDB, Graphiti, Mem0 | Semantic caching, hybrid Graph RAG retrieval, and agent memory in Inference, Data and Knowledge Fabric |
| Evaluation & benchmarking | DeepEval, Promptfoo, Braintrust, Arize Phoenix, LangSmith, RAGAS; τ²-bench, Terminal-Bench, and SWE-bench Verified as external baselines | Regression testing and multi-judge pipelines in Eval Fabric |
| LLM observability & FinOps | OpenTelemetry (GenAI semantic conventions), Langfuse, Helicone, Arize Phoenix | Token, latency, and cost telemetry in Inference Fabric |
Data Engineering & Lakehouses
Local-first for development: prototype on embedded engines, then verify SQL semantics, catalog support and data contracts on each target cloud engine. The compatibility criteria are in Emerging Tech.
| Role | Tools | Use / evaluation scope |
|---|---|---|
| Processing engines | Polars, DuckDB, Apache Arrow, Apache DataFusion, Apache Spark (PySpark), Databricks, Pandas | Vectorized execution in Data Fabric; local-first verdict in Emerging Tech |
| Low-latency analytics | ClickHouse 26.8 LTS | Event-query candidate in Streaming Fabric; validate ingestion, duplicates and cost before adoption |
| Streaming & CDC | Apache Kafka 4.3 (KRaft) / Redpanda, Apache Flink 2, Apache Fluss, Debezium 3, Google Pub/Sub, Dataflow | Streaming ingestion in Data Fabric and the client engagements in About |
| Orchestration | Dagster, Apache Airflow 3.3; Cloud Composer / MWAA (version-specific support), Prefect 3 | Asset-aware batch/streaming pipelines described in Collaboration |
| Storage, table formats & catalogs | Apache Iceberg v3, DuckLake, Delta Lake, Lance, Apache Parquet, Apache Polaris, Unity Catalog, Lakekeeper, Cloud Storage, Amazon S3 | Open-table-format integration in Data Fabric |
| Transformation & modeling | dbt Core (v2 release candidate under evaluation), Fusion, SQLMesh, SQL, Kimball dimensional modeling, Data Vault 2.0, data contracts (ODCS) | Reproducible dbt diagnosis and patch trials in Engineering Fabric |
Cloud & Platform Infrastructure
Infrastructure described as code, with policy checks before changes. Cross-cloud portability requires provider-specific designs and validation.
| Role | Tools | Use / evaluation scope |
|---|---|---|
| Public & private cloud | Google Cloud Platform (GCP), Amazon Web Services (AWS); OpenShift / OpenStack as private references | Reference architectures in Production Blueprints |
| Infrastructure as Code | Terraform / OpenTofu (modular landing zones), Terragrunt, Crossplane, OPA Rego, Conftest, Kyverno | Multi-cloud foundations in Production Blueprints; policy-as-code spike in Emerging Tech |
| Containers & orchestration | Docker, Kubernetes (GKE / EKS; DRA, Gateway API Inference Extension), Google Cloud Run (GPU), Amazon ECS / Fargate | Containerized runtimes across Active Labs |
| GitOps & secrets management | ArgoCD, HashiCorp Vault / OpenBao | GitOps reconciliation and zero-trust credentials in Platform Fabric |
| Platform observability | Prometheus, Grafana, Mimir, Loki, Tempo, Pyroscope, Grafana Alloy | Centralized SRE telemetry in Fabric Ops |
| CI/CD & local toolchain | GitHub Actions, just, uv, pre-commit | The toolchain this site runs on |
Languages, Databases & Standards
| Role | Tools | Use / evaluation scope |
|---|---|---|
| Programming languages | Python (free-threading only where profiled), SQL, Rust, Scala, Bash | Rust/Python router experiments in Inference Fabric; language choice follows the workload |
| Databases | PostgreSQL, AlloyDB, DuckDB, ClickHouse, MySQL, Redis, BigQuery, Snowflake, Redshift | Target compatibility evaluated in Emerging Tech |
| Engineering standards | Data contracts (ODCS), schema evolution, idempotency, OpenTelemetry semantic conventions, pre-commit quality gates, Conventional Commits, SemVer | Applied across every project, including this repository |
| Tool UX/UI | Zensical, semantic HTML, token-based CSS, native JavaScript, Playwright, axe-core | This site as a progressive-disclosure and accessibility case; UX/UI principles |
Version review · 2026-09-06: DuckDB 1.5 is the stable line; 2.0 is an alpha planned for October. dbt Core v2 reached release candidate 1 on 2026-09-02. Airflow 3.3, Kafka 4.3 and ClickHouse 26.8 LTS each need their own compatibility matrix. ADK 2.0 workflows do not imply SDK feature parity. Sources and adoption conditions are in the dated radar.
For upstream developments, adoption conditions and proposed tests, see Emerging Tech. For workload-based comparisons of these tools and alternatives, see the Blog, grouped by architectural role with reviewed licenses and operating constraints.