# Swarnajit Roy — AI/ML Systems Engineer & Researcher > Swarnajit Roy is an AI/ML Systems Engineer and Researcher based in Bengaluru, India. He specializes in Retrieval-Augmented Generation (RAG) systems, computer vision ensembles, automated analytics pipelines, and applied machine learning research. He is currently pursuing his B.Tech in Electronics & Computer Science Engineering (AI & ML) at Christ (Deemed to be University) with a CGPA of 3.55 / 4.00. ## Professional Summary - **Primary Role**: AI / ML Engineer & Systems Developer - **Location**: Bengaluru, Karnataka, India - **Education**: Bachelor of Technology in Electronics & Computer Science Engineering (AI & ML), Christ (Deemed to be University), Aug 2023 - May 2027 (CGPA: 3.55 / 4.00) - **Key Capabilities**: RAG Systems, Vector Search (FAISS), Computer Vision Ensembles (PyTorch, EfficientNet, ViT), Automated EDA Pipelines, Quantum Hybrid Steganography, Signal Processing (ObsPy STA/LTA). - **Email**: royswarnajit341@gmail.com - **Phone**: +91 62893 02303 - **GitHub**: https://github.com/coder-royswarnajit - **LinkedIn**: https://www.linkedin.com/in/swarnajit-roy-7a4879286 - **Résumé**: https://swarnajitroy.in/assets/Swarnajit_Roy_Resume.pdf - **Portfolio Architect & Designer**: M S Rishav Subhin (`msrishav-28`) — https://github.com/msrishav-28 --- ## Portfolio Architecture & Engineering Credits - **Lead Web Architect & UI/UX Designer**: M S Rishav Subhin (`msrishav-28`) - **Email**: ms.rishav289@gmail.com - **Phone**: +91 98355 97637 - **Location**: Bengaluru, Karnataka, India - **ORCID iD**: https://orcid.org/0009-0002-7431-9666 - **Social & Professional Profiles**: - Personal Portfolio: https://portfolio-msrishav.vercel.app/ - GitHub: https://github.com/msrishav-28 - LinkedIn: https://linkedin.com/in/m-s-rishav-subhin - X (Twitter): https://x.com/msrishav_28 - Hugging Face: https://huggingface.co/msrishav - Peerlist: https://peerlist.io/msrishav289 - Product Hunt: https://www.producthunt.com/@m_s_rishav_subhin - Dev.to: https://dev.to/msrishav-28 - Hashnode: https://hashnode.com/@msrishav-28 - ORCID: https://orcid.org/0009-0002-7431-9666 - **Architectural Contribution**: Engineered SwarnajitDex v2.0 design system, Web Audio SFX sound engine, live GitHub telemetry engine, and responsive UI framework. - **Repository**: https://github.com/msrishav-28/swarnajitroy-portfolio --- ## Technical Stack & Capabilities - **Machine Learning**: PyTorch, Scikit-learn, Hugging Face, XGBoost, OpenCV, ObsPy - **LLM & RAG**: RAG Architecture, FAISS Vector Indexing, Prompt Design, IBM Watsonx.ai, Groq API, LangChain - **Languages**: Python, SQL, Java, C++ - **Delivery & Tools**: Streamlit, FastAPI, Docker, Git, REST APIs, Linux - **Domains**: Retrieval-Augmented Generation, Deepfake Detection, Predictive Analytics, Steganography, Planetary Seismology --- ## Experience Record ### 1. UST — AI/ML Intern - **Period**: Apr 2026 – Jul 2026 (Remote) - **Summary**: Developed an internal AI Auto-EDA & Modelling Pipeline in Streamlit. Automated dataset ingestion, statistical profiling, grounded LLM explanations via Groq API, and baseline model comparisons, reducing initial dataset review time from days to hours. ### 2. Minu Travels — AI Chatbot Developer Intern - **Period**: Oct 2025 – Nov 2025 (West Bengal, India) - **Summary**: Designed and implemented a production RAG system processing 100+ business documents with automated chunking, FAISS vector embeddings, and confidence-based fallback logic. ### 3. AICTE - Edunet Foundation — AI & ML Intern - **Period**: Jun 2025 – Jul 2025 (Remote) - **Summary**: Designed end-to-end regression ML prediction workflows, data preprocessing pipelines, feature engineering, and MAE-based model evaluation interfaces. ### 4. IEEE Computer Society, Bangalore — AI Research Intern - **Period**: Apr 2025 – Jul 2025 (Karnataka, India) - **Summary**: Architected a deepfake detection ensemble combining EfficientNet, ResNet, and Vision Transformer (ViT) backbones with weighted sampling for class imbalance. --- ## Case Studies & Systems 1. **Tekshila** (Flagship · Developer Tools) - AI documentation and code-quality assistant delivering annotated feedback directly inside GitHub pull requests. - Stack: LLM APIs, GitHub API, Python. - Repository: https://github.com/msrishav-28/Tekshila 2. **AI Auto-EDA & Modelling Pipeline** (Flagship · Analytics Pipeline) - UST Global analytics workflow for profiling, visualization, explanation, and baseline model selection. - Stack: Python, Streamlit, Scikit-learn, Groq API, Pandas, Plotly. 3. **Deepfake Detection System** (Flagship · Computer Vision) - Research ensemble combining EfficientNet, ResNet, and Vision Transformers for real-vs-fake video detection. - Stack: PyTorch, EfficientNet, ResNet, Vision Transformer, OpenCV. - Repository: https://github.com/coder-royswarnajit/Deepfake-Detection 4. **Agragrati** (Lab · LLM & RAG) - Resume intelligence prototype for skill extraction, structured candidate-role relevance scoring, and job API discovery. - Stack: Python, Streamlit, Groq API, External Job APIs, Docker. - Repository: https://github.com/coder-royswarnajit/Agragrati 5. **Battery Life Estimation** (Lab · Predictive ML) - SMAMIA 2025 research study evaluating single and stacked regressors via a Battery Degradation Factor (R² ≈ 0.98, MAE ≈ 1.5h). - Stack: Python, Scikit-learn, XGBoost, Pandas. - Repository: https://github.com/coder-royswarnajit/Battery-Prediction 6. **Mentora** (Lab · LLM & RAG) - Document-learning assistant using IBM Watsonx.ai to turn educational PDFs into grounded summaries and active-recall quizzes. - Stack: Python, Streamlit, IBM Watsonx.ai. - Repository: https://github.com/coder-royswarnajit/Mentora 7. **Multimodal Summarizer** (Lab · NLP) - Unified Poke-Summarizer interface for documents, audio, video, and live news powered by Groq inference. - Stack: Python, Streamlit, Groq API, Media Pipelines. - Repository: https://github.com/coder-royswarnajit/Poke-Summarizer 8. **Planetary Seismic Detection** (Lab · Analytics Pipeline) - Interactive signal-processing toolkit using ObsPy to analyze STA/LTA ratio triggers across Mars and Moon datasets. - Stack: Python, ObsPy, Streamlit. - Repository: https://github.com/coder-royswarnajit/SesmicDetection --- ## Research Publications & Awards - **Quantum Hybrid Steganography Model** (*Presented Aug 2025*): Published at International Conference on Sustainable Analytics and AI (SaaAI '25). Achieved 88.06% security strength and SSIM ~ 0.996. - **Comparative Analysis of ML Models for Battery Life Estimation** (*Conference Paper Dec 2025*): Published at International Conference on Smart Materials, Advanced Manufacturing & Industrial Applications (SMAMIA '25). Stacked ensemble achieved R² ~ 0.98. - **Winner** — Gen'AI Work-a-thon, Christ University. - **First Place** — Extra-Curricular Activities Award, CUESTIC EC Club (2024–2025). - **Leadership**: Wing Head — Skill Development & Training, Industry Institute Interaction Cell (IIIC), Christ University. --- ## Optional Full Details For un-truncated architectural specs and code snippets, see [llms-full.txt](https://swarnajitroy.in/llms-full.txt).