👋 Hi, I'm Sumit Kumar

🚀 AI Engineer | ML Research Engineer | Full Stack Developer

---

🎯 About Me

class AIEngineer:
    def __init__(self):
        self.name = "Sumit Kumar"
        self.role = "AI/ML Engineer & Full Stack Developer"
        self.education = "IIT Madras - BS in Data Science & Programming"
        self.location = "Bengaluru, India"
        self.expertise = [
            "Generative AI & LLMs",
            "Computer Vision & 3D Reconstruction", 
            "Backend Architecture & APIs",
            "MLOps & Cloud Deployment"
        ]
        
    def current_focus(self):
        return {
            "🔬 Research": "Multimodal AI & Time Series Forecasting",
            "🏗️ Building": "AI-Powered Production Systems",
            "📚 Learning": "Advanced Agent Architectures & Graph Neural Networks",
            "🌟 Goal": "Transforming AI Research into Scalable Solutions"
        }

🔹 AI Engineer passionate about building impactful solutions in Computer Vision, NLP, and Generative AI
🔹 Skilled in designing end-to-end ML systems: RAG chatbots, 3D vision models, multimodal AI
🔹 Experienced in backend API development, microservices, and cloud-based MLOps
🔹 Thriving on turning complex AI research into production-ready solutions that drive real-world impact

---

📊 GitHub Statistics

---

💼 Professional Experience

🏢 Current Roles.

🔹 AI Engineer Intern 📍 ForeignAdmits (VisaMonk AI) | Bengaluru 📅 July 2025 – Present Built FA-Admission Backend with Node.js, Express.js, MongoDB Developed AI-powered University Chatbot with RAG pipeline & FAISS Created AI Document Processing System using Tesseract OCR & GPT-4 Engineered Email Outreach Platform with automated AI content generation

🔹 AI/ML Research Engineer 📍 Freelancer | Remote (South Korea) 📅 Oct 2025 – Present Multimodal Emotion Recognition: 92.53% accuracy on IEMOCAP Fashion Trend Forecasting with ensemble models (N-BEATS, PatchTST) Research on Graph Neural Networks & adaptive modality weighting MLOps pipelines with uncertainty quantification

🎯 Recent Positions

🔹 Backend Developer (Freelancer) | ElitCeler Technologies | Aug 2025 - Oct 2025

• Architected RESTful APIs for 2 full-scale e-commerce platforms
• Built Bazar Story & Printrove WMS backends with 50+ endpoints
• Integrated AWS S3, Shopify OAuth, payment gateways

🔹 AI & Data Science Intern | HTS Tech Solutions | Mar 2025 - Jul 2025 | PPO Received ✅

• YOLOv11-based rust detection for cell towers: 85% accuracy
• 3D Model reconstruction using OpenMVG/OpenMVS
• Reduced model build time from 12 hours → 3-4 hours
• Report delivery time: 3 days → <24 hours

🔹 Product and AI (Freelancer) | Arfve | Stockholm, Sweden | Apr 2025 - Jul 2025

• AI agent-driven lead generation & automation
• UX improvements & prototype features for accelerator cohort

🔹 Full-Stack Developer | Devvoy | Jan 2025 - May 2025

• AI-powered therapy platform with LLM-driven dialogues
• Voice-enabled interactions using React, FastAPI, ElevenLabs TTS
• Mentored 3+ contributors on Git workflows and deployment

---

🛠️ Tech Stack

💻 Languages

🌐 Backend & APIs

🤖 AI/ML & LLMs

🛢️ Databases

☁️ Cloud & DevOps

🎨 Frontend

---

🚀 Featured Projects

🎤 Hemut Voice AI - Trucking Dispatch Automated trucking dispatch system with AI-powered voice calls Tech: FastAPI, PostgreSQL, Vapi.ai, Twilio Impact: 90% reduction in manual dispatch operations Features: AI-driven voice conversations Real-time webhook processing International call support Driver management APIs	🔍 AI-Powered Code Analysis Platform Comprehensive code quality assessment across 10+ languages Tech: FastAPI, Google Gemini AI, FAISS, MongoDB Features: RAG engine for codebase Q&A AST parsing for security vulnerabilities GitHub integration Real-time progress tracking.
📞 VoxFlow.ai - Voice Calling Platform Scalable AI-powered voice calling platform Tech: Node.js, Twilio, Supabase, Groq, Deepgram Features: RESTful APIs with RBAC Job queues for campaign management Speech-to-text transcription WebRTC integration	💬 AI-Powered RAG Chatbot Full-stack chatbot with vector search and real-time processing Tech: Python, FAISS, Redis, Flask Highlights: 90% accuracy with hybrid RAG Web scraping & data indexing Multi-tenant deployment TTS generation
🏭 Fault Detection & Quality Inspection CNN-based defect detection for manufacturing Tech: Keras, Flask, OpenCV, Node.js Results: 93% detection accuracy 18x faster inspection time 7,000+ training images	🦀 3D Rust Detection for Cell Towers YOLOv11 + 3D reconstruction pipeline Tech: YOLOv11, OpenMVG/OpenMVS, Node.js Achievements: 85% detection accuracy 12 hrs → 3-4 hrs model build time 3 days → <24 hrs report delivery

📂 View More Projects | 📊 Data Science Projects

---

🏆 Achievements & Certifications

🥇 Top 3 in Industrial AI Solutions Hackathon (2024)
📰 Published Machine Learning research in IIT Madras Newsletter (Nov 2024)
👥 Led 200+ students programming community with Codeforces/LeetCode challenges
🎓 BS in Data Science & Programming from IIT Madras (2024-2027)
💼 PPO Received from HTS Tech Solutions (2025)

---

📈 Contribution Graph

---

🎯 Core Competencies

Domain	Skills
🤖 Generative AI	RAG Architecture, Multi-Agent Systems, Prompt Engineering, Function Calling, LangChain, LlamaIndex
🧠 Machine Learning	CNNs, Transformers, Graph Neural Networks, YOLOv11, LoRA/QLoRA Fine-tuning, Multimodal AI
🏗️ Backend Engineering	REST APIs, GraphQL, Microservices, JWT/OAuth, RBAC, API Gateway, Rate Limiting
☁️ MLOps & Cloud	Model Deployment, Drift Detection, AWS SageMaker, Docker/Kubernetes, CI/CD Pipelines
📊 Data Engineering	ETL Pipelines, Apache Spark/Kafka, Vector Databases, Big Data Processing
🔧 System Design	Scalable Architecture, Load Balancing, Caching Strategies, Database Optimization

---

💡 What I'm Currently Working On

🔬 Research:
  - Multimodal Emotion Recognition with Graph Neural Networks
  - Fashion Trend Forecasting using Ensemble Time Series Models
  - Adaptive Modality Weighting for Robust AI Systems

🏗️ Building:
  - AI-Powered Voice Communication Systems
  - Document Processing Pipelines with OCR & LLM Integration
  - Scalable RAG Architectures for Enterprise Applications

📚 Learning:
  - Advanced Agent Architectures (ReAct, Reflexion)
  - Real-time Streaming AI Applications
  - Production MLOps Best Practices

---

📫 Let's Connect!

💬 Open to collaborations on AI/ML projects, research opportunities, and interesting tech challenges!

---

🌟 "Transforming AI Research into Production-Ready Solutions" 🌟

⭐ If you find my work interesting, feel free to star my repositories! ⭐