Real-Time Query Agent Built on Llama 3 and CrewAI

Added Jun 2025 3 design docs

Hosted AI APIs are convenient, but many teams need alternatives: open-weight models avoid per-token costs, keep data in-house, and can run where commercial APIs cannot. Developers therefore need to know how to stand up an agent on an open model, and how orchestration frameworks behave when the model underneath is not a commercial black box. That integration skill is what this project builds. The intern connects Llama 3 to CrewAI to create an agent that answers user queries in real time. Working in Python, the intern configures CrewAI to use the open-weight model as its reasoning engine, defines an agent with a clear role and instructions, and builds the query loop through which user questions are dispatched to the agent and answers streamed back. Along the way the intern handles the practical differences that come with open models, including prompt formatting expectations, response parsing, and latency management, and documents how the agent's behavior compares with the same setup on a hosted model. Though deliberately compact, the project teaches foundations with lasting value: wiring an orchestration framework to a specific model backend, understanding where model choice changes agent behavior, and gaining confidence with the open-source side of the AI stack that many production teams now depend on.

Related projects

Sandbox for Building and Benchmarking Agentic AI Workflows

Agentic AI systems are notoriously hard to evaluate: an agent that shines on one task fails quietly on another, and teams need somewhere to build, test, and compare agents rigorously before trusting them with real work. This project builds that environment. The intern creates an experimentation platform where agents built with CrewAI and Langchain, powered by OpenAI models, can be constructed, configured, and benchmarked side by side. The system supports defining test scenarios, running agents against them, and capturing results for comparison, turning subjective impressions of agent quality into measurable outcomes. Agent customization tools let users adjust prompts, tools, and collaboration patterns, while user profile management keeps each experimenter's configurations and history organized. A FastAPI backend orchestrates runs and a Streamlit frontend, all written in Python, makes experimentation interactive and repeatable. The project builds a skill the AI industry urgently needs, evaluation engineering: designing harnesses that reveal how agent systems actually behave. The intern finishes with both a working testing platform and demonstrated judgment about what makes agents reliable.

educationresearchcrewaifastapi+4

View project

Hard

Academic Literature Review Automation With Multi-Agent Pipelines

Researchers and graduate students drown in publications: keeping up with a single field can mean scanning hundreds of papers a month, and literature reviews consume weeks of focused effort. This project builds a platform where autonomous agents shoulder that reading load. The intern implements a multi-agent pipeline with CrewAI and Langchain in which specialized agents collect academic literature, analyze it, and produce structured summaries, with LlamaIndex indexing papers for retrieval and OpenAI models powering advanced summarization. A FastAPI backend coordinates agent runs and stores results, while a Streamlit dashboard lets researchers define topics, monitor collection progress, and read synthesized findings, with personalization features tuning the platform to each user's research interests over time, all built in Python. The project demands real engineering judgment, including orchestrating agents that hand work to one another, grounding summaries in source documents to limit fabrication, and presenting machine-read literature in a form scholars trust. Completing it demonstrates the intern can apply agentic AI to knowledge work at research grade, a capability valued in both academia and industry R&D.

educationresearchcrewaifastapi+5

View project

Medium

Autonomous Literature Review Agent for Academic Paper Analysis

A literature review is the most repetitive stage of research: find candidate papers, read abstracts, discard the irrelevant, extract methods and findings from the rest, and synthesize the field's state. Researchers and graduate students spend weeks on this cycle, and the manual approach does not scale as publication volume grows. An agent that runs the cycle autonomously, while keeping the human in control of direction, is genuinely valuable. The intern builds that agent as a Python system served through FastAPI. LangChain orchestrates the research loop, planning searches, deciding which papers merit deeper reading, and chaining analysis steps, while LlamaIndex indexes the gathered papers so extraction and synthesis are grounded in actual text. The system works with both OpenAI APIs and Llama models, letting the intern compare hosted and open-weight performance on the same tasks. A user profile layer personalizes output depth and focus areas for each researcher, and an experimentation playground lets users adjust the agent's strategy and observe how its behavior changes. Finishing the project demonstrates autonomous agent design, dual-framework retrieval-augmented generation, multi-model integration, and the evaluation discipline needed to verify that an agent's summaries faithfully reflect the papers they cite.

educationresearchfastapilangchain+4

View project