Abhijeet Jha
AI/ML Engineer | Building LLM, Agentic AI & ML System
Ready for work
Abhijeet is ready for their next project!
The Great AI ROI Lie Nobody in wants to Say Out Loud
0
1
Jarvis CLI is a Python-based voice assistant for Windows that enables hands-free desktop interaction through voice commands and an animated graphical interface. The assistant listens for a wake word, interprets spoken requests, and performs practical actions such as launching local applications, opening websites, searching Google or YouTube, playing media, retrieving the current time, and handling basic system commands.
The project uses a modular architecture that separates speech processing, text-to-speech, command routing, desktop actions, and the GUI layer. It includes an animated orb interface that visually represents assistant states such as idle, listening, speaking, and code-generation activity. The voice and conversational pipeline was further explored with Google Gemini integration for transcription and AI-generated responses, while the project also contains local speech-recognition work based on Vosk.
I designed and developed the assistant’s voice-command workflow, desktop automation capabilities, modular Python architecture, animated GUI, speech-processing integration, command normalization logic, and Gemini-powered conversational enhancements. The project demonstrates practical work in voice interfaces, AI assistant design, desktop automation, and human-friendly interaction flows.
0
5
SuperAgent is a multi-agent orchestration system built to automate complex, multi-step workflows across research, planning, coding, and general problem-solving tasks.
The platform uses a hierarchical Hub-and-Spoke architecture powered by LangGraph and Google Gemini. A central Supervisor agent interprets the user’s goal, breaks the task into smaller steps, and routes work to specialized agents based on capability.
The system includes a Coder Agent with isolated local code execution, a Generalist Agent for research and ideation, and persistent shared state to maintain context across agent handoffs. A React-based management dashboard provides visibility into agent activity, workflow progress, and approval-required actions.
Safety is a core part of the design. SuperAgent applies iteration limits to control runaway execution and requires human approval before sensitive code execution steps proceed. The code-execution workspace is isolated through Docker-based containerization.
I designed and developed the agent architecture, orchestration flow, safety model, code-execution environment, persistent context layer, backend services, and management dashboard.
0
30
A modern AI infrastructure tool that estimates GPU VRAM requirements for LLM inference, fine-tuning, and quantization. It helps ML engineers configure a model workload, understand where memory is consumed, and identify suitable GPU hardware before deployment.
0
41