Elena Artemova

Projects with local language models

Examples working Local AI in action: from hybrid search to data analysis artificial intelligence

Each project is a real-world scenario using local LLMs for business tasks. All solutions run within your controlled infrastructure, so data does not need to be sent to external AI services.

Discuss your scenario

Below are working projects that can be adapted to your company's needs. Each lists the business value, technologies used, and readiness status.

01

Protocol_1.0 — transcripts, meeting minutes, and Obsidian export

A ready product for meeting and audio processing: transcription with diarization, decision and task extraction, Markdown and Obsidian export. Everything runs locally — data never leaves your environment.

Demo available
Highlights
  • Transcription and diarization (speaker separation)
  • Automatic extraction of decisions, tasks, and deadlines
  • Meeting summaries and a searchable minutes archive
  • Markdown export with direct Obsidian integration
Business value

Reduces the time needed to prepare meeting minutes and helps preserve decisions and tasks. All data stays on your machine.

02

Automation with Obsidian and hybrid search

A scheduled job monitors a folder, creates a local knowledge graph in Obsidian, and enables hybrid search (vector and keyword) across all materials.

Demo available
Highlights
  • A scheduled tracker automatically processes new files
  • Local knowledge graph in Obsidian with cross-links
  • Hybrid search: semantic + full-text
  • Chat interface to query the graph
Business value

Turns scattered documents into a connected knowledge base that answers natural-language questions. Reduces search time from hours to seconds.

04

Database queries via Text-to-SQL

A local AI agent translates natural-language questions into SQL queries, executes them, and returns answers. Everything runs locally — no data leaves your infrastructure.

Demo available
Highlights
  • Natural questions → SQL without developer help
  • PostgreSQL, SQLite, MySQL support
  • Local execution keeps data within the configured environment
  • Query review and confirmation before execution
Business value

Enables business users to work with data without SQL knowledge. Reduces load on analysts and DBAs. Security ensured by local execution.

05

Email sorting and draft response generation

A local AI sorts incoming email by topic and urgency, classifies requests, drafts contextual responses based on the company knowledge base, and flags critical messages for manual review.

Demo available
Highlights
  • Automatic classification of incoming email
  • Context-aware draft response generation
  • Urgent and critical message detection
  • Uses corporate templates and the knowledge base as sources
Business value

Reduces email processing time by 40–60%, lowers the risk of missing important messages, standardizes response quality.

06

Automatic document categorization and tagging

An agent monitors a shared folder, categorizes new documents by content type, extracts key metadata, and organizes files into a structured directory.

Demo available
Highlights
  • Real-time folder monitoring
  • Document classification by type and content
  • Extraction of dates, amounts, parties, and other metadata
  • Automatic sorting into folder structure
Business value

Eliminates manual document sorting, accelerates search, reduces document loss risk. Immediate ROI in high-volume document departments.

07

Lead Qualifier

Scores B2B leads using an open formula, shows the score, and leaves the deal decision to a person. The scoring criteria are transparent and configurable.

Demo available
Highlights
  • An open scoring formula shows how the score is calculated
  • Transparent criterion-by-criterion review for every lead
  • A person always makes the deal decision
  • Integration with CRM and lead sources
Business value

The sales team spends time only on leads worth pursuing; the scoring criteria are clear and do not become a black box.

08

Procurement Planner

Builds a demand forecast and a purchase plan with quantities, budget, and visible confidence per item.

Demo available
Highlights
  • Demand forecast based on consumption history
  • Purchase plan: quantity, budget, and timing
  • Visible confidence for every item
  • Inventory and seasonality accounted for
Business value

Purchasing is planned from data, not memory: the budget and quantity for each item are visible in advance, together with forecast confidence.

09

Company Data Consultant

Answers questions about company or website data and honestly says “I don’t know” when the fact is unavailable. Uses only approved sources.

Demo available
Highlights
  • Answers based strictly on uploaded company or website data
  • An honest “I don’t know” instead of a fabricated answer
  • A source link for every answer
  • Local deployment in a closed environment
Business value

Customers and employees get accurate answers from current data, while the company avoids reputational risk from fabricated facts.

10

CustDev Agent

Researcher: turns interviews and notes into a map of observations, needs, and testable product opportunities.

Demo ready for presentation
Highlights
  • Separates direct quotations from interpretations
  • Stages: preparation → coding → synthesis → artifact
  • Findings linked to primary excerpts and evidence strength
  • A reproducible research brief instead of scattered notes
Business value

The team receives a report, quotations, and an opportunity map for the next hypothesis test; a person makes and confirms critical decisions.

11

Super Oracle

Solution engineer: guides a product task through five engineering methods in 75 minutes and returns five artifacts and a plan. It runs locally, so the data never leaves your computer.

Demo ready for presentation
Highlights
  • Root-cause map across six categories
  • Resolving a contradiction without compromise
  • Finding uncontested market space
  • Risk and priority scoring in numbers
Business value

Instead of debate and intuition, a system turns the task into a plan: a cause map, contradiction resolution, strategy, risk list, and priorities.

12

Muse Agent

Critic: checks the interface, user flow, and copy for clarity, accessibility, consistency, and honesty toward the user.

Demo ready for presentation
Highlights
  • Dark patterns and manipulative mechanics
  • Interface accessibility review
  • Observations with evidence and priorities
  • An honest alternative for every issue
Business value

The product is reviewed before release: observations, evidence, and fix priorities replace subjective opinions.

First step

Choose a ready-made solution or describe your process.

Pick one of the twelve agents and subagents, or fill in the questionnaire for custom development.

Choose an agent Fill out the questionnaire