Autonomous agents and multi-agent workflows that plan, use tools, and coordinate to solve complex tasks. Built with LangChain, LangGraph, and local language models.
Single agents that plan, use tools, and complete multi-step tasks without constant human input — for web, data, and code.
Graph-based workflows (LangGraph) where specialised agents collaborate — each with its own scope, memory, and tools.
Agents that navigate websites, fill forms, scrape data, and interact with web interfaces using Playwright.
Retrieval-Augmented Generation pipelines. Agents that query vector stores and answer questions over private documents.
Private, offline AI using Ollama. No API costs, no data leaving your machine — full control over the model.
Agents that read, write, and execute code — for automated analysis, legacy code migration, and report generation.
Breaking the goal into subtasks. Defining what tools (search, code exec, APIs) each agent needs access to.
Mapping the workflow as a directed graph — which agents coordinate, what triggers handoffs, and how state flows.
Implementing each agent node, writing precise system prompts, and testing tool-calling behaviour in isolation.
Running the full pipeline on real tasks, fixing failure modes, adding fallbacks, and logging for observability.
Local LLM-driven agent that automates browser tasks and multi-step web workflows.
Multi-agent workflow coordinating interviews for 1,000+ recruiters with calendar and email tool integration.
Graph-based multi-agent framework that analyses legacy codebases and orchestrates refactoring agents.