AI Engineer | AI & Robotics Enthusiast
Building multimodal AI systems with LLMs, Speech AI, and Computer Vision.
I specialize in applying LLMs, voice interfaces, and vision models to build intelligent, real-world AI systems. My work spans agentic automation, offline AI assistants, and cognitive architectures that combine perception, reasoning, and adaptive behavior.
Currently pursuing M.Tech in AI & Data Science at IIT Patna, where I’m exploring multimodal human–AI interaction systems.
- Human Visual Perception & Cognitive Response System: A Bio-Inspired Architecture for Machine Perception, Understanding, and Learning
- IIT Patna capstone project integrating vision, LLM reasoning, episodic memory, and RL to build a bio-inspired cognitive loop.
- Tech: PyTorch · OpenCV · MCP · RL · Multimodal AI
- AIPAT: AI Powered Assistance Tool for Terminal.
- Features: Shell automation, AI code generation, chat, and more.
- Powered by Meta's LLaMa model with local processing.
- ASTRA: An ethical malware project built for testing and cybersecurity education.
- Technologies: Python (Socket Programming, Websockets)
- Multi-platform support for Windows, Linux, and macOS.
- LocalTalkingAssistance: — Offline Voice AI Assistant
- Wake-word → VAD → Whisper ASR → LLM → XTTS.
- End-to-end offline voice assistant with natural speech interaction.
- Tech: Whisper · XTTS · Audio Pipeline · Python
- AI Intern @ IBM SkillsBuild
- Learned and explored the AI fundamentals.
- Completed courses on Project Management and Placement Assistance.
- Cybersecurity Intern @ Tutelr Infinity
- Gained exposure to DFIR, penetration testing, and blue teaming strategies.
- Network Security Intern @ CyberDojo
- Implemented and managed pfSense firewalls and Snort IDS/IPS.
- Simulated network architectures for hands-on learning.
- Python
- Django | DRF | React.js | WebSockets | Socket | Next.js | Flask | PyTorch | LangChain | HuggingFace | FastAPI | NumPy | Pandas | Matplotlib
- LLM Orchestration, Agentic Workflows, Memory Systems, Prompt Engineering, n8n, MCP
- Wazuh | TheHive | MISP | ELK Stack | pfSense | OPNsense | Snort IDS/IPS | Metasploit | Nmap
- Git | Docker | Nginx | AWS | GitHub Actions | Ansible | Terraform
- Linux (Ubuntu/Kali) | Windows | macOS
- Cognitive architectures for multimodal AI
- Agentic pipelines for task automation
- Real-time speech + vision fusion systems
- 📧 Email: someshs.ce@gmail.com
- 🌐 Portfolio: someshs.tech
- 💼 LinkedIn: LinkedIn Profile
Feel free to explore my repositories and projects. I’m always open to collaboration, feedback, and exciting discussions about technology. Let’s build something amazing together! 🚀
