Project Overview

As AI tooling became central to modern development workflows, I built a suite of production AI agents designed specifically to automate and enhance the repetitive cognitive overhead of day-to-day developer work. Each agent is purpose-built with multi-step reasoning pipelines, tool orchestration, and MCP-style API integrations — going well beyond simple prompt wrappers.

Agents Built

Git Commit Agent

• Analyses current staged code changes using a diff-aware context pipeline and generates well-structured, conventional commit headlines and detailed messages.
• Understands code semantics — distinguishing between a refactor, a bug fix, a feature addition, and a performance improvement — producing commit messages that are informative rather than generic.
• Reduces time spent on commit descriptions from minutes to seconds while improving message quality and team communication.

Daily Status / Standup Agent

• Generates concise, accurate daily standup reports from the developer's committed code changes and chat history.
• Summarises what was completed, what is in progress, and any blockers — formatted for direct use in team standups or Monday.com status updates.

Code Review Agent

• Reviews code locally before push, flagging security vulnerabilities, code smells, logic errors, and performance anti-patterns — effectively acting as a pre-PR code reviewer.
• Structured to catch the same class of issues that GitHub Copilot code review would surface, but running locally on demand before the PR is even opened.

Monday.com Task Generator Agent

• Reads a user story or ticket description alongside the current codebase context, then generates granular micro dev tasks via the Monday.com API — including file paths, functions to modify, and acceptance criteria.
• Bridges the gap between product-level user stories and developer-level implementation tasks, reducing planning overhead significantly.

Resume Builder Agent

• AI-powered resume optimisation agent that analyses job descriptions, extracts ATS keywords, and suggests targeted improvements to resume content and structure.
• Generates tailored resume variants aligned to specific job roles without losing factual accuracy.

Technical Architecture

• Multi-Step Reasoning Pipelines: Each agent uses structured agentic workflows with multi-step reasoning — not single-shot prompts — enabling complex analysis and contextual decisions.
• Tool Orchestration: Agents call external tools (Git CLI, Monday.com API, file system) as part of their reasoning loop, following MCP-style integration patterns.
• Model Flexibility: Built to support both OpenAI (GPT-4) and Google Gemini models, with a switching layer to use the most cost-effective model per task type.