AI · LLMs · Agents · Research

AI Projects

by Praneeth Vadlapati

Skills used: Gen AI, LLMs, observability, evaluation, LangChain, LangGraph, Agentic AI, AI Engineering, and many more

AI
LLMs
Python
LangChain
LangGraph
LangSmith

Projects: 31
Building since: 2023
Open-source PRs: 4

~/projects

❯ ls --sort=date | head -4
pepip/                 Apr 2026
agent-action-guard/    Apr 2026
tool-see/              Dec 2025
index-rag/             Nov 2025
❯ ls | wc -l
31 projects · 2023 → today
❯

Latest projects

Recent flagship builds — shipped as open-source packages, research and tools.

Oct 2026

Newest · Agent Runtime

Agent-RT: Fast, lightweight runtime for production AI agents

A provider-neutral runtime for building production AI agents in Python and TypeScript. It provides a clear execution boundary for tools, streaming, structured output, permissions, approvals, memory, checkpoints, sandboxing, observability, and deployment without requiring a heavyweight orchestration stack.

Includes built-in production controls such as capability grants, execution budgets, guardrails, sandbox boundaries, resumable checkpoints, vector-database switching, and optional Agent Action Guard integration.

GitHub Website Docs PyPI NPM

Dec 2025 – Apr 2026

Flagship · AI Safety

Agent Action Guard: Safety Layer for Autonomous AI Agents

Developed a research-driven safety framework for autonomous AI agents that detects and prevents harmful actions before execution. Introduces a novel Agent Safety Benchmark, labeled dataset, and a deep learning-based action classifier built using PyTorch and NLP techniques.

The system classifies agent actions in real time as safe or harmful, enabling enforcement of data privacy, security policies, ethical constraints, and regulatory compliance. It is designed as a model-driven guardrail layer for LLM-powered and tool-using agents.

GitHub Website PyPI Package NPM Package

Apr 2026

Dev tooling

pepip: Python version of pnpm — Shared package-version store

pepip keeps an immutable shared package-version store and symlinks each project's .venv to the exact versions it resolved. This significantly reduces disk usage and speeds up installs for package-heavy Python workflows, especially AI/ML projects.

GitHub Website PyPI Package

Nov 2025

RAG · Citations

Index-RAG: Storing Text Location in Vector Databases for QA tasks

Generates questions that answer a chunk, and stores embeddings with text location for precise citations. Uses Python, OpenAI, embeddings, RAG with PineCone vector database, and FastAPI. Enables fine-grained, citation-accurate QA by indexing embeddings with exact source text offsets, improving trust and explainability.

View on GitHub

Open-Source Contributions

My contributions to popular open-source projects of other organizations.

Aug 2025
traceloop/openllmetry

#3330 #3336 docs #100

LLM Telemetry, OpenAI API
Sep 2025
BrainBlend-AI/atomic-agents

#177

Agents, MCP, tools, prompts, OpenAI API

Project archive

Every build since 2023 — research experiments, agent frameworks, RAG systems and LLM apps.

2025

iRAT: Replanning and Controlled Retrieval for Robust LLM Reasoning

Used re-planning and controlled retrieval to gain accuracy over CoT and RAG by implementing prompting, Transformer-based LLMs, NLP, evaluation and integration of LLM capabilities. Authored a research paper. Uses Python, OpenAI API, Transformers, a fine-tuned LLM, NLP, CoT, reasoning, RAG, and APIs. Demonstrates superior reasoning accuracy via structured replanning and retrieval control using re-rankers.

July 2025

Production-ready template for large Generative AI projects

Streamlines the development and deployment of large-scale Generative AI projects, ensuring best practices in code structure and documentation. Accelerates enterprise GenAI adoption by standardizing architecture, scalability, and maintainability.

Dec 2025

Reasoning Router: Dynamic Multi-Strategy Reasoning with LangGraph

Leverages LangChain and LangGraph to dynamically route reasoning tasks to the most appropriate LLMs’ reasoning strategy based on user prompts. Uses LangSmith for LLM observability, monitoring. Enhances reasoning quality and cost efficiency by adaptive strategy selection at inference time.

Oct 2025

Data Classifier: An AI-driven approach to Label LLM Training Data

Classifies large datasets into relevant and irrelevant data using NLP and Transformers, creating a pure dataset for LLM training or fine-tuning.

Oct 2025

Smart-Prompt-Eval: Evaluating LLM Robustness using Manipulated Prompts

Evaluates LLMs using manipulated prompts to stress-test the robustness of LLMs using Python. Systematically stress-tests LLMs to uncover failure modes beyond standard benchmark evaluations.

Sep 2025

2024

UpdAgent: AI Agent Version Control Framework for Real-Time Updation of Tools

A framework that enables versioning for agents, which allows the updation of versions.

Nov 2024

AutoTrendyKeywords: Real-Time AI-Driven Trend-Based SEO Using LLM

Trend-based SEO to make the content stay relevant to the users despite rapid changes in trends.

Oct 2024

LML-DAP: Language Model Learning a Dataset for Data-Augmented Prediction

Instead of ML, the paper uses the process of an LLM learning a dataset for context-aware classification decisions based on relevant data like RAG. Outcome: The predictions were accurate and transparent, as the system generates the reason behind each prediction.

Oct 2024

AGIent and Tool-Mart: A Super-Agent Framework Enabled by Marketplace-Driven Tool Selection

Enables dynamic set of tools for agents using a database.

Sep 2024

AutoPureData: Automated Filtering of Undesirable Web Data to Update LLM Knowledge

Filters unwanted text using LLMs and uses Vector DB and Retrieval-Augmented Generation (RAG).

July 2024

AutoWatcher: a Real-Time Context-Aware Security Alert System using LLMs

Uses multimodal LLMs to monitor camera footage and create alerts on detection of suspicious threats.

May 2024

The Power of Roles: Investigating Impact of Three Message Types on LLM Responses

Investigates the impact of three message roles, "system," "user," and "assistant," on responses.

Mar 2024

AutoPrecisePrompts: Automated LLM-based Prompt Engineering for Data Processing

Automated Prompt Engineering and testing to ensure perfect prompts are crafted.

Jan 2024

2023

Agent-Supervisor: Supervising Actions of Autonomous AI Agents to Ensure Ethical Compliance

Flags harmful actions by agents before execution.

Dec 2023

Hierarchical Management of AI for Automated Monitoring and Query Resolution

Utilizes a large LLM to answer complex queries and supervise a smaller LLM.

Oct 2023

Does Seed Matter?: Investigating the Effect of Random Seeds on LLM Accuracy

Investigates the influence of the seed value that is used to control the randomness of an LLM.

Aug 2023

Investigating the Impact of Linguistic Errors of Prompts on LLM Accuracy

Investigates the impact of errors such as grammatical errors and spelling errors on LLM accuracy.

Jun 2023

iSearch: Enhanced Intelligent Search Based on a Query using LLMs

Automates the Prompt Engineering of hard prompts and the testing process to ensure perfect prompts are crafted.

Apr 2023

Multilingual Prompting in LLMs: Investigating the Accuracy and Performance

Since the training data of LLM is mostly in English, the experiments finds the accuracy of various tasks using prompts in different languages.

Feb 2023

LangChain
Applications

AI Projects

Latest projects

Agent-RT: Fast, lightweight runtime for production AI agents

Agent Action Guard: Safety Layer for Autonomous AI Agents

pepip: Python version of pnpm — Shared package-version store

Index-RAG: Storing Text Location in Vector Databases for QA tasks

ToolSEE: Agent Tool Search Engine for Efficient and Scalable Tool Discovery using Retrieval

InvesTutor-RAG: An Intelligent Financial Tutor using Retrieval-Augmented Generation

Open-Source Contributions

Project archive

iRAT: Replanning and Controlled Retrieval for Robust LLM Reasoning

Production-ready template for large Generative AI projects

Reasoning Router: Dynamic Multi-Strategy Reasoning with LangGraph

Data Classifier: An AI-driven approach to Label LLM Training Data

Scratchpad 3Q Reasoning: Surfacing Assumptions to Mitigate Hallucination and Improve Truthfulness in Large Language Models

Smart-Prompt-Eval: Evaluating LLM Robustness using Manipulated Prompts

GameNet3-AI: Decentralized social platform for gamers

UpdAgent: AI Agent Version Control Framework for Real-Time Updation of Tools

AutoTrendyKeywords: Real-Time AI-Driven Trend-Based SEO Using LLM

LML-DAP: Language Model Learning a Dataset for Data-Augmented Prediction

AGIent and Tool-Mart: A Super-Agent Framework Enabled by Marketplace-Driven Tool Selection

AutoPureData: Automated Filtering of Undesirable Web Data to Update LLM Knowledge

AutoWatcher: a Real-Time Context-Aware Security Alert System using LLMs

The Power of Roles: Investigating Impact of Three Message Types on LLM Responses

AutoPrecisePrompts: Automated LLM-based Prompt Engineering for Data Processing

Agent-Supervisor: Supervising Actions of Autonomous AI Agents to Ensure Ethical Compliance

Hierarchical Management of AI for Automated Monitoring and Query Resolution

Does Seed Matter?: Investigating the Effect of Random Seeds on LLM Accuracy

Investigating the Impact of Linguistic Errors of Prompts on LLM Accuracy

iSearch: Enhanced Intelligent Search Based on a Query using LLMs

Multilingual Prompting in LLMs: Investigating the Accuracy and Performance

LLMs - Chat with a Wikipedia page

LLMs - Feedback summarizer

LLMs - Chat with an image

LLMs - Prompt shortener

LLMs - Agents

LLMs - Chat with Data

LLMs - Model Deployment

Sentiment Analysis