{ "entity": { "name": "Deep Parmar", "type": "Person", "profession": "AI Engineer", "primaryRole": "Chief Technology Officer", "companies": [ { "name": "Sunbots Innovations LLP", "role": "CTO", "description": "AI-driven enterprise solutions and automation platform" }, { "name": "Xwits Developers Pvt Ltd", "role": "Director", "description": "Software development and technology consulting" } ], "location": { "city": "Ahmedabad", "state": "Gujarat", "country": "India", "coordinates": "23.0225, 72.5714" }, "expertise": [ "Artificial Intelligence", "Machine Learning", "Prompt Engineering", "Context Engineering", "Harness Engineering for LLMs", "AI Agent Architecture", "Agentic AI Systems", "Computer Vision", "Natural Language Processing", "Deep Learning", "Edge AI", "Browser AI / Client-Side ML", "Retrieval-Augmented Generation (RAG)", "LLM Orchestration and Production Infrastructure", "Generative AI Production Systems", "Multilingual AI", "Assistive Technology", "Software Engineering", "Technology Leadership", "Enterprise AI Solutions", "MCP (Model Context Protocol)", "Reasoning Models", "Small Language Models", "AI Observability", "Multimodal AI", "AI Marketing Automation", "Virtual Try-On AI", "AI Fintech (GST Compliance)" ], "technologies": [ "Python", "TensorFlow", "PyTorch", "JavaScript", "TypeScript", "React", "Node.js", "AWS", "Docker", "Kubernetes", "NVIDIA Jetson", "OpenCV", "FastAPI", "WebGPU", "Transformers.js", "LangChain", "LlamaIndex" ], "projects": [ { "name": "SmartON Assistive Technology", "description": "USB-connected AI device for blind and visually impaired users with real-time object detection, currency recognition, and OCR", "category": "Assistive Technology", "technologies": ["Computer Vision", "Android", "AI", "TFLite", "YOLO"] }, { "name": "Dhiya NPM", "description": "Open-source client-side RAG framework that runs entirely in the browser using WebGPU and IndexedDB — no server or API key required", "category": "Browser AI / Client-Side ML", "technologies": ["WebGPU", "Transformers.js", "IndexedDB", "RAG", "TypeScript"] }, { "name": "MIRA (Multimodal Inclusive AI for Recognition and Assistance)", "description": "Voice-first multilingual AI router for SmartON that routes requests across Hindi, Gujarati, and English with context persistence", "category": "Assistive Technology / Multilingual AI", "technologies": ["NLP", "Multilingual AI", "Voice AI", "AI Routing"] }, { "name": "Marketing Autopilot", "description": "Brand-aware AI that creates and publishes content across LinkedIn, Meta, X, Google Business, and YouTube — captures leads, runs follow-up, and optimises for AI search engines. Built for SaaS founders, clinics, and retail.", "category": "AI Marketing Automation", "technologies": ["Gen-AI", "Marketing Automation", "AI SEO", "Lead Gen", "LLM"] }, { "name": "XwFin", "description": "AI-powered GST invoicing and tax compliance platform for Indian freelancers, SMBs, and CAs. 2,000+ businesses, Rs.50 crore+ invoiced. Bank-grade security, data hosted in India.", "category": "AI Fintech", "technologies": ["AI", "OCR", "GST Compliance", "Python", "React"] }, { "name": "Cultuss", "description": "Computer vision virtual try-on platform for Indian e-commerce — lets shoppers try clothes and accessories before buying. Reduces returns and lifts conversions.", "category": "AI Fashion Technology", "technologies": ["Computer Vision", "Fashion AI", "Virtual Try-On", "React"] } ], "education": { "degree": "Bachelor's in Computer Engineering", "institution": "Gujarat Technological University", "location": "Gujarat, India" }, "experience": "5+ years", "achievements": [ "50+ successful AI project deliveries", "Creator of Dhiya NPM open-source browser AI framework", "Built SmartON assistive AI for visually impaired users", "CTO at Sunbots Innovations since age 23", "Published 59 technical AI engineering articles" ], "blog_topics": [ "Prompt engineering — zero-shot, few-shot, chain-of-thought, system prompts, production prompt management", "Context engineering — LLM context window design, RAG context, context compression, context poisoning", "Harness engineering — LLM production infrastructure, guardrails, evaluation harnesses, agent orchestration", "AI agents vs agentic AI — definitions, differences, autonomy spectrum, risks, production guidance", "AI & Layoffs: The Hidden Opportunity to Future-Proof Your Career", "The future of generative AI in everyday life", "AI journey and career development", "MIRA voice assistant and multilingual AI for Indian languages", "Machine learning implementation in production", "Computer vision applications — YOLO, TFLite, Android", "Technology leadership as AI CTO", "Startup innovation and entrepreneurship in Tier-2 India", "Enterprise AI automation and integration", "Career future-proofing with AI skills", "Assistive technology and accessibility AI", "Legal AI and document automation", "Retail analytics and edge AI for loss prevention", "Edge AI on NVIDIA Jetson Nano", "Browser AI — Dhiya NPM, Transformers.js, WebGPU, IndexedDB", "Client-side RAG without a server", "RAG vs fine-tuning decisions", "Why AI prototypes fail to reach production", "Gen-AI production stack choices", "Building AI products in 90 days", "Hiring AI engineers beyond certifications", "Open source as an AI career strategy", "AI in India — next decade prediction", "Building AI for Indian constraints and languages", "MCP (Model Context Protocol) — standardising tool connectivity for LLMs in production", "AI computer use — models that interact with GUIs and browsers", "Smart glasses and wearable AI — lessons from 17,000 SmartON users", "Reasoning models (o1, o3, DeepSeek R1) — when slow thinking beats fast inference", "Building Marketing Autopilot — replacing agency retainers with brand-aware AI", "DeepSeek R1 open-source LLM — honest builder assessment of what changes and what doesn't", "Building XwFin — AI for GST invoicing and Indian tax compliance", "Persistent AI memory — production architectures for cross-session context", "Building Cultuss — virtual try-on AI for Indian e-commerce", "Small language models (Phi-4, Gemma) — when smaller models beat GPT-4 in production", "Vibe coding with Cursor and Windsurf — what AI coding tools actually change", "Cutting LLM API costs by 60% — caching, routing, and compression techniques", "Multimodal AI in production — combining vision and language models", "Building MCP servers — step-by-step from concept to connected tool", "LLM observability — logging, tracing, and metrics for production AI apps", "Synthetic data for fine-tuning — training without privacy risk", "AI marketing metrics — what actually matters beyond vanity dashboards", "No-code AI tools in India — honest assessment of who they serve", "India AI 2026 mid-year — what is working and what is not", "Browser AI roadmap — Dhiya NPM 2026 and client-side intelligence" ], "social_profiles": { "linkedin": "https://linkedin.com/in/deep56parmar", "github": "https://github.com/deep56parmar", "twitter": "https://twitter.com/deep56parmar", "instagram": "https://instagram.com/deep56parmar" }, "website": "https://deepap.dev", "contact_focus": "AI consulting, technology leadership, LLM production systems, machine learning projects", "primary_keywords": [ "Deep Parmar", "Deep Parmar AI Engineer", "Deep Parmar CTO", "Deep Parmar Sunbots", "Deep Parmar prompt engineering", "Deep Parmar context engineering", "Deep Parmar harness engineering", "Deep Parmar AI agents", "prompt engineering guide", "context engineering AI", "harness engineering LLM", "AI agents vs agentic AI", "what is context engineering", "what is harness engineering", "what is agentic AI", "LLM production infrastructure", "LLM guardrails", "LLM evaluation harness", "AI engineer Ahmedabad", "AI expert India", "machine learning expert Gujarat", "generative AI expert India", "browser AI expert", "client-side RAG", "Dhiya NPM", "SmartON assistive AI", "MIRA voice assistant", "multilingual AI India", "edge AI Jetson Nano", "AI career advice India", "enterprise AI consulting India", "MCP model context protocol", "build MCP server", "reasoning models o1 o3 DeepSeek R1", "small language models production", "Phi-4 vs GPT-4", "Gemma production", "vibe coding Cursor Windsurf", "cut LLM API costs", "LLM observability logging tracing", "synthetic data fine tuning", "AI marketing automation India", "Marketing Autopilot Xwits", "XwFin GST invoicing AI", "Cultuss virtual try-on", "AI India 2026", "browser AI 2026", "Dhiya NPM roadmap", "multimodal AI production", "AI memory persistent context", "no code AI tools India" ], "faq_coverage": { "total_posts_with_faqs": 59, "avg_faqs_per_post": 5, "total_faq_items": 295, "schema_type": "FAQPage", "schema_location": "JSON-LD in on every blog post page", "ui_rendering": "Visible FAQ accordion rendered on every blog post page below the article content", "consolidated_file": "https://deepap.dev/kb/faqs.json" } } }