name: Devashish Bose
role: AI Engineer | LLM Applications | Full-Stack Developer
education: B.Tech, Internet of Things (IoT)
institute: Madhav Institute of Technology & Science, Gwalior
location: India
current_focus:
  - LLM post-training workflows
  - Retrieval-Augmented Generation (RAG)
  - Agentic workflows and evaluation
  - GATE CS 2027 preparation
working_style: learn-by-building

I build practical software at the intersection of AI engineering, data, and product development. My work includes document retrieval, LLM-enabled workflows, machine-learning applications, and local-first web products. I enjoy understanding the complete path—from data and model behavior to APIs, interfaces, evaluation, and deployment.

• 🧠 LLM Engineering: prompt design, structured outputs, SFT/RLHF workflows, and application integration.
• 📚 Retrieval Systems: embeddings, semantic search, vector stores, retrieval evaluation, and RAG pipelines.
• 📊 Machine Learning & Analytics: data preparation, feature engineering, model evaluation, and visualization.
• ⚙️ Software Engineering: Python APIs, TypeScript applications, databases, and end-to-end product workflows.
• 🔬 Continuous Learning: GATE CS 2027, ML foundations, agent orchestration, reliability, and safety.

• LLM application development and prompt engineering
• Retrieval-Augmented Generation (RAG) and hybrid retrieval concepts
• Embeddings, semantic search, and vector databases
• LangChain, Hugging Face, ChromaDB, and FastAPI
• Retrieval evaluation and structured response workflows
• SFT/RLHF workflow support, annotation, and post-training tasks

• Python, Pandas, NumPy, Scikit-learn
• Data cleaning, EDA, feature engineering, and model evaluation
• Random Forest, classification, forecasting, and time-series workflows
• SQL, R, Power BI, Tableau, Hadoop, and Hive
• Matplotlib and Seaborn for analysis and visualization

• FastAPI, Flask, Node.js, and REST APIs
• TypeScript, JavaScript, React, Next.js, and Django
• PostgreSQL, Prisma, ChromaDB, Redis, and IndexedDB
• Git, GitHub, Docker, Vercel, and deployment workflows
• Local-first architecture, PWA patterns, and offline-friendly experiences

_{Additional tools and platforms: LangChain, Hugging Face, ChromaDB, embeddings, semantic search, Pandas, NumPy, Power BI, Tableau, Hadoop, Hive, Prisma, and IndexedDB.}

Selected builds across retrieval, AI-assisted workflows, personal systems, and applied machine learning.

A document question-answering system built with FastAPI, LangChain, embeddings, and ChromaDB. The project explores document ingestion, chunking, vector retrieval, and answering over retrieved context. Its recorded dense-retrieval evaluation includes Recall@1: 0.78, Recall@5: 1.00, Recall@10: 1.00, and MRR: 0.8867.

An AI-assisted job-search workflow concept focused on bringing job discovery, resume parsing/tailoring, and outreach support into one experience. The project has explored LLM provider integration, search tooling, and email workflow considerations.

A personal learning and career operating system designed to coordinate study plans, track actual effort, manage revision, and adapt work to available capacity. It combines a Next.js PWA interface with local-first data and planning workflows.

A local-first fitness and nutrition PWA for recording workouts, food, body metrics, and habits. Its architecture emphasizes offline use and local persistence, with synchronization concepts for connected use.

• LLM & RAG: retrieval design, embeddings, evaluation, and grounded answer generation.
• Agents & Orchestration: tool calling, task decomposition, agent memory, and workflow coordination.
• AI Operations: latency, cost, monitoring, reliability, and evaluation.
• Safety & Ethics: hallucination reduction, privacy, bias, and prompt-injection awareness.
• Foundations: machine learning, data structures, algorithms, and GATE CS 2027 preparation.

_{Statistics are rendered by third-party services and may be temporarily unavailable or delayed.}

“Understand the fundamentals. Build the system. Measure what matters. Improve continuously.”

I value clear thinking, practical experimentation, readable engineering, and products that solve real problems. I’m interested in collaborating on AI/ML, data, and software projects where thoughtful implementation and continuous learning matter.

⚡ Thanks for visiting — let's build something useful. ⚡