Sfoglia categorie
Esplora
Fiverr Pro
Italiano
$
USD
I will build a personal ai assistant that knows your notes docs and decisions
Amit Patel
Informazioni su questo servizio
A personal AI that actually knows your stuff. Point Claude at your Obsidian vault, Notion export, or PDF/Word folder. Ask it anything get answers grounded in your own writing, not the internet.
You keep notes but never re-read them. You've written the same insight three times because you forgot the first two. You want to talk to your knowledge base like a colleague. RAG is the fix retrieval-augmented generation over your data.
Deliverables:
- ChromaDB vector store indexing your notes/docs with embeddings
- SQLite knowledge graph capturing decisions, tags, cross-refs
- Claude (or your preferred LLM) queried with your context injected
- Nightly enrichment: auto-tag, summarise, link
- Chat interface: CLI, web UI, or Obsidian plugin integration
- Runbook: how to add new docs, tune retrieval, back up
Why me:
I run this exact stack on my own vault every day 700+ notes, 23 architecture decision records, all indexed and queryable by Claude across sessions. Custom subagents, session-capture hooks, MCP integration. Same pattern, applied to your knowledge.
Not a wrapper around ChatGPT real RAG with vector search and knowledge graph.
Message me first every job starts with a 15-min scoping
Scopri di più su Amit Patel
Amit Patel
Engineering Manager
- DaRegno Unito
- Membro daset 2026
- Tempo di risposta medio1 ora
Lingue
Inglese, Gujarati
I am an Engineering Manager with over 15 years of experience in FinTech, FX, and payments. My expertise lies in enhancing operational efficiency and fortifying security through a blend of investigative skills and AI-augmented delivery. I am dedicated to continuous improvement and problem-solving to ensure system reliability and client satisfaction.
Altri servizi della categoria Sviluppo AI offerti da me
FAQ
What input formats do you support?
Markdown (Obsidian, notes.app export), PDF, DOCX, TXT, HTML, Notion export, Confluence export. If it's text or extractable text, it works.
How much data can it handle?
Basic up to 500 docs, Standard up to 2000 docs, Premium unlimited (with tuning for large corpora — chunking, filtering, hybrid retrieval).
Which LLM does it use?
Claude (Anthropic API) by default — best quality for RAG. OpenAI, Gemini, or local via Ollama supported on Premium.
Where does it run?
Locally on your machine, or self-hosted on your server. Nothing leaves your environment except relevant chunks + prompts to the LLM API of your choice.
What about privacy?
Your docs stay local. Only relevant retrieved chunks sent to the LLM per query. Full audit log of what was sent so you can review.
Can I use Obsidian?
Yes — Obsidian vaults are the best-fit input format. I can also deliver as an Obsidian plugin integration on Premium so you query directly from the editor.
How do I add new docs later?
Watch folder auto-ingests new files, or manual re-index command. Runbook covers both plus how to rebuild the whole index if needed.
Can it write things, not just answer?
Yes — it can draft answers, summarise, generate outlines from your existing notes, produce first drafts of new documents grounded in your knowledge. Not just Q&A.
What about MCP integration?
Premium includes an MCP server exposing the knowledge base as a Claude Code tool — you can query it directly from any Claude Code session.
What's out of scope?
Building your knowledge base from scratch (you need existing notes/docs). Fine-tuning a model on your data (different technique). Multi-user enterprise deployment (see other gigs).
