AI/ML Enthusiast | AWS JAM 2026 Runner-Up | Amazon ML
School ’25 Alumnus
Creator of Zeni and YogAI, building
intelligent systems with LLMs, computer vision, and voice
AI.
B.Tech Computer Science Student at Graphic Era Hill
University (2023–2027)
The future is shaped by intelligent systems, and I’m passionate about building AI solutions that create real impact.
Currently working on projects like Zeni and YogAI, focused on ultra-low-latency voice AI, robot control, and real-time learning.
Winner of 2 hackathons and Amazon ML School ’25 alumnus, dedicated to advancing AI and machine learning.
You can find my projects and updates here, or get in touch via the Contact section.
For more, check out my Resume, Projects, and Skills.
Bachelor of Technology
Specialization: CSE (Hons.) with Machine
Learning and Artificial Intelligence
GPA: 8.65
Aug 2023 – Jul 2027
Completed Intermediate with 83.40% and
Matriculation with 82.17%.
July 2021 – July 2023
Issuer: DeepLearning.AI
Issued: Apr 2025
Credential ID: G66Y716CLQNC
Issuer: IBM
Issued: Nov 2024
Credential ID: C3DREIU3V7EO
Issuer: DeepLearning.AI
Issued: Nov 2024
Credential ID: 4X3THBJG0HSQ
Issuer: University of Michigan
Issued: Sep 2024
Credential ID: 4XEFBM5EJKV8
Secured 2nd place among 400+ participants in a competitive cloud challenge organized in association with Amazon Web Services (AWS).
Developed YogAI posture correction system using AI and computer vision, winning ₹40,000 prize at Neural Nexus Hackathon organized by Graphic Era University.
Created FaultExpert AI/ML fault diagnosis model detecting production line defects at Graphic Era Hill University hackathon.
Completed Amazon Machine Learning School 2025, gaining advanced hands-on AI and ML skills.
WeCode | Aug 2025 – Present ()
Mentoring students in AI/ML at my college. Conducted sessions on machine learning and AI basics, helping others grasp complex concepts simply.
Artificial Intelligence (AI) and Machine Learning
Personal Client Project | Aug 2025 – Present ()
Currently building a comprehensive fitness app that integrates AI-powered Yoga posture detection, personalized gym routines, and nutritional guidance — all in one platform. The app uses computer vision, voice AI, and real-time feedback to deliver a tailored wellness experience.
Technologies: Computer Vision, LLMs, Python, TensorFlow
Zeni is a production-grade, ultra-low-latency voice AI assistant deployed as a physical AI receptionist at Graphic Era Hill University. Built by a team of CSE students, it features a ~500ms first-word latency, achieving a highly responsive full-duplex conversational experience seamlessly in Hindi and English.
Technical Highlights:
Technologies: Llama 3.3 70B, FastAPI, Google Cloud ASR/TTS, ChromaDB, WebSockets, Android/Java
“The Zeni speaks so realistically and so fast, it feels like you’re talking to a real human.”
—Student User, GEHU Bhimtal
“This project has real product potential — it’s awesome and ready to be launched.”
—Head of the Department of Computer Science, GEHU
An AI-powered bilingual attendance system using face recognition via CCTV or webcam to automate attendance in real-time.
Built with YOLOv8, DeepFace, TensorFlow, and OpenCV, this project captures and identifies faces from live feeds or videos, applies Region of Interest (ROI) filtering, and logs attendance directly into CSV — all with high speed and accuracy.
[Technologies: YOLOv8,
DeepFace, OpenCV, Numpy, Pandas]
[GitHub:
DeepTrack Repo
⭐ Loading stars...]
Shankar Singh is an AI researcher, developer, and engineer with deep interests in large language models (LLMs), computer vision, and real-world applications of artificial intelligence. His work bridges the gap between cutting-edge AI and its practical deployment through mobile and web applications.
He is the creator of several notable open-source projects, including DeepTrack, a real-time facial recognition attendance system using YOLOv8 and DeepFace, and AlphaMind, an offline bilingual campus assistant built using LLaMA 3.2 and a RAG pipeline. His projects span voice interfaces, offline AI systems, and multimodal applications combining speech, vision, and LLMs.
Shankar is an alumnus of the prestigious Amazon ML Summer School 2025, and has won multiple hackathons, including the Neural Nexus AI/ML International Hackathon and the 24-Hour TechFest Hackathon at Graphic Era Hill University.
His recent freelance work involves building an integrated fitness and wellness platform combining gym routines, AI-driven yoga posture correction, and personalized nutrition tracking.
For more about his work, visit his GitHub, check out his LinkedIn, or explore his open-source DeepTrack project.
My work revolves around developing AI systems that operate in real-world environments — from educational institutions to physical spaces like gyms or classrooms. I specialize in building practical, offline-first applications that integrate local LLMs, computer vision, voice interaction, and retrieval-augmented generation (RAG).
The common theme across my projects is solving real, on-ground problems using modern AI. Whether it's AlphaMind — an offline bilingual campus assistant, DeepTrack — a facial recognition attendance system, or YogAI — a real-time yoga posture corrector, each system aims to deliver fast, accurate, and accessible experiences with minimal hardware requirements.
cont→
cont→
I build with a strong focus on privacy (offline/local models), inclusivity (Hindi-English bilingualism), and edge-efficiency (optimized pipelines with low latency). My stack includes tools like FastAPI, LLaMA via Ollama, YOLOv8, DeepFace, FAISS, ChromaDB, AWS (EC2, S3), and Docker.
Over the last few years, I’ve worked across diverse AI domains — from face recognition and posture correction to building multilingual AI assistants and personalized recommendation engines. Many of my projects originate from everyday problems I encounter as a student or developer — whether it's automating classroom attendance, understanding human behavior through Reddit activity, or making offline AI tools accessible in Hindi and English.
While most of my work is open-source and individually developed, I’ve also collaborated in hackathons and team projects, often winning or placing at national-level events. I value hands-on, end-to-end execution — from data pipelines and model training to deployment and optimization. Some systems, like DeepTrack and Zeni, required integrating multiple complex modules, including LLMs, YOLO-based object detection, TensorFlow models, and speech recognition—all working together in a high-performance setting.
I approach AI from a builder's mindset, but I’m also drawn to the ethics, generalization, and long-term impacts of deploying such systems at scale. A recurring question in my mind is: Can we make AI systems that are both useful and contextually aware, without requiring massive cloud infrastructure or exposing private data?
Some of my other public projects include:
My work continues to evolve, and I’m currently experimenting with lightweight RAG-based tools, emotion-aware models, and agents that can reason with voice, text, and image inputs — all while remaining fully offline or edge-deployable.
I'm open to new collaborations, freelance projects, or just tech chats.
Email:
shankarbisht1224@gmail.com
LinkedIn:
linkedin.com/in/shankarsingh077
GitHub:
github.com/shankarsingh077
Feel free to reach out — whether it's a project, a question, or just a hello.
DeepTrack — Real-time AI Attendance via Face Recognition
Zeni — Real-Time Bilingual Voice AI Robot
RedditUserPersona — LLM-based Persona Generation from Reddit