Suchith Koduru

AI Engineer building intelligent systems.

I work on LLM agents, multi-agent architectures, and memory-augmented AI at 10xscale.ai. Previously research at IIT Delhi. I care about building AI that's reliable, contextual, and useful in production.

Suchith Koduru

About

I'm an AI Software Engineer at 10xscale.ai, where I build multi-agent systems using LangChain, LangGraph, and FastMCP. I've contributed to the open-source agent framework Agentflow — PRs #17 and #19 are merged — and built an MCP server for PostgreSQL that lets agents query structured data with natural language.

I design persistent memory systems using Qdrant and Mem0 for long-horizon reasoning, and I prototyped a Raspberry Pi-powered robot with real-time AI-generated responses for live demos. Before this, I was a research intern at IIT Delhi under Prof. Kapil Tomar, where I built a full-stack Maternal Health Risk Predictor — benchmarking SVM, Random Forest, XGBoost, and Gradient Boosting to reach 99% accuracy.

I graduated from IIT Delhi in 2025 with a B.Tech in Computer Science. I'm interested in making AI agents more capable, contextually aware, and reliable for production use.

Work

Featured

Agentflow Contributions

Contributed to the open-source AI agent framework with focus on code structure, unit test coverage, and PR management. Built a robust foundation for scalable agent development.

  • Python
  • LangChain
  • Git
  • Unit Testing
Featured

Deep Research Agent

Built a deep research agent with web searching capabilities for Hugging Face's MCP Hackathon. Autonomously researches topics, synthesizes information, and provides comprehensive answers.

  • Python
  • MCP
  • LangChain
  • HuggingFace
Featured

AI-Powered Robotic Prototype

Designed and built a Raspberry Pi-powered robot with real-time AI-generated responses for live demonstrations. Integrates LLM inference with physical hardware for interactive human-robot conversations.

  • Python
  • Raspberry Pi
  • LLM APIs
  • Hardware

Other Projects

MCP PostgreSQL Server

Model Context Protocol server enabling natural language querying of PostgreSQL databases with agent integration.

  • Python
  • FastMCP
  • PostgreSQL

Customer Churn Prediction

End-to-end ML pipeline for telecom churn prediction using Random Forest with SMOTE, SHAP interpretability, and threshold optimization.

  • Python
  • Scikit-learn
  • SHAP

CIFAR-100 Classification

High-confidence image classification using WideResNet with Sharpness-Aware Minimization and temperature scaling.

  • PyTorch
  • WideResNet
  • SAM

Background

I work primarily with Python, LangChain, LangGraph, PyTorch, and PostgreSQL. I deploy with Docker on Linux, and have experience with C++, Flask, JavaScript, Qdrant, FastMCP, and Scikit-learn.

B.Tech in Computer Science, Indian Institute of Technology Delhi (2021–2025).

JEE Advanced 2021: AIR 479 (of 1.2 lakh). JEE Mains 2021: AIR 620 (of 10 lakh+). TS-EAMCET 2021: Rank 201 (of 2.2 lakh).