A plugin for your agentic framework that optimizes code using the GEPA algorithm (Genetic-Pareto LLM-driven search).

Synth Python SDK for Managed Research, Research Factory, and GEPA/GELO optimizer workflows.

Claude Code for DSPy: Comprehensive CLI to Optimize Your DSPy Code. our AI-Powered DSPy Development Assistant

CLI text optimizer built on GEPA. Uses Agentic Coding CLI's as mutator and observer -- no api keys required

Allen is an open-source AI engineering team that runs on your repos — specialist agents, deterministic orchestration through workflows, and approval gates you control. Drive it from chat, Slack, or Linear.

Optimize any AI agent’s skills, tools/MCP, and prompts against your own evals.

GEPAzilla: open-source GEPA prompt optimizer with datasets, scorers, and telemetry.

Local code search for AI agents: six fast, purpose-built tools that return ranked answers, not raw grep. Because maybe grep isn't all you need... 🍬

Evolving agent harnesses: a research program on how far N orchestrated calls of a small model can rival a frontier model. We evolve the harness (structure + prompts) with reflective optimizers + a verified-acceptance gate.

GEPA and GELO optimizer runbooks, SDKs, and hosted optimizer surfaces for Synth.

Self-evolve Gemini CLI instructions, commands, and skills via the gemini CLI itself — GA + GEPA/DSPy, with hard gates before apply.

Production-ready boilerplate for building and automatically optimizing LangChain RAG applications. Implements three-layer architecture: Build (LangChain) → Measure (MLflow) → Optimize (GEPA + MEGA).

A benchmark, alignment pipeline, and LLM-as-a-Judge for evaluating the clinical impact of ASR errors.

Prompt optimisation with GEPA: mine a compliance rubric from labelled decisions. 30% more violations caught, starting from a one-line prompt.

A physics-grounded, agent-driven digital twin for HP Metal Jet S100 3D printer

LLM agents that generate, verify, and evolve Triton GPU kernels. Includes a reward-hack-resistant benchmarking harness with strict correctness verification and fresh-input evaluation. Achieves up to 174.7× over PyTorch eager and outperforms FlexAttention (1.48×) and SDPA (1.17×) on selected workloads.

Reproducible benchmark: prompt optimization (DSPy GEPA) vs fine-tuned small models on real, human-annotated data. Every number rerunnable.

A brief experiment applying GEPA (optimize_anything) to automatically compress Python solutions on code.golf.

DSPy-Powered Prompt & Code Optimization via the Model Context Protocol (MCP).

Budge is the experimentation platform for agents.