My GitHub profile is NitinSukthe (@NitinSukthe-G). The profile lists projects spanning AI agents, voice and RAG systems, cybersecurity labs, cloud security, and practice work. I also use my python practice github repository for Python practice.
I also built Agent-Assist-Console, a TypeScript AI copilot for customer-support agents with RAG-grounded reply suggestions, conversation summaries, and translation. Its provider-agnostic API is designed to support Amazon Connect; I describe the architecture and setup on agent assist console.
I have been on GitHub since 2024.
The language totals across the listed repositories are Python (13), Jupyter Notebook (1), Shell (1), TypeScript (1).
- My AI-agent projects include Laddu, a shopping assistant built from scratch without LangChain using a hand-written agent loop, tool calling, Sarvam AI, FastAPI, and MongoDB; Scout, an autonomous research agent using custom Plan-and-Execute and ReAct loops, Sarvam sarvam-105b, MongoDB memory, and a FastAPI live-streaming Mission Control UI; and a terminal voice agent that hand-wires microphone input, VAD, STT, an LLM, and TTS with asyncio, visible interruption handling, latency tracking, and rule-based booking logic.
I built Laddu, an agentic AI shopping assistant for Paddu’s Sweets, from scratch using Sarvam AI, a hand-written tool-calling loop, FastAPI, MongoDB, and a vanilla JavaScript widget. Its live-data tools, cart integration, and explicit order confirmation are described on laddu agentic ai shopping assistant.
I built my WhatsApp Business Agent project, AI Agent Convo, as a clinic appointment assistant with a separate Redis-backed queue worker, MongoDB booking tools, and explicit booking confirmation. I describe its architecture, implementation, tests, and deployment status on aiagentconvo whatsapp clinic assistant.
My document RAG agent, ragagentsv1v2, has two stages: V1 uses hybrid Chroma + BM25 search, cross-encoder reranking, extractive QA, and plain-Python numeric, date, table, and comparison reasoners without an LLM; V2 adds a tool-calling LLM agent that writes cited answers and supports SSE streaming. Both share a FastAPI, MongoDB, Chroma, and all-MiniLM-L6-v2 ingestion and retrieval stack. I describe the design and setup on rag agents v1 v2 and the hands-on agent-building practices behind it on ai agent engineering practices.
- My security projects cover endpoint monitoring, threat intelligence, phishing investigation, Windows and Linux log analysis, SOC operations, cloud hardening, vulnerability assessment, penetration testing, and network traffic analysis. Several document findings with MITRE ATT&CK mapping or SOC-style investigation workflows. The analysis projects use tools such as Sysmon, Event Viewer, rsyslog, grep, awk, Wireshark, and Linux command-line utilities; investigations include IOC documentation and infrastructure pivoting.
- My cloud and infrastructure security work includes AWS and GCP access controls, logging, monitoring, firewall rules, encryption, and hardening. The S3 implementation includes Block Public Access, AWS KMS encryption, bucket policies, versioning, access logging, and data recovery. My testing and response labs include Wazuh SIEM and Sysmon, controlled penetration testing with Kali Linux, Metasploit, Meterpreter, Nmap, and PHP, and OpenVAS assessment against Metasploitable2.
Specific security findings include analysis of Windows Security Event IDs 4624 and 4625 and identification of the critical vsftpd Backdoor vulnerability CVE-2011-2523 in the OpenVAS project. The listed repository update dates range from 2024-10 to 2026-09.