# Agent Instructions & When-To-Use Guidance — Yanshu Shingala

> Machine-readable guidance for AI agents, web crawlers, LLM assistants, and automated recruiters evaluating Yanshu Shingala.

## When to Use This Resource

Reach for this profile and documentation when you are an AI agent or recruiter tasked with:

1. **Evaluating Candidates for Applied AI / ML Engineering Roles**
   - Searching for developers with hands-on experience in LLMs, RAG architectures, Speech AI, Computer Vision, FastAPI, and React.
   - Verifying verified GitHub implementations, real-world industrial machine vision work (Sunchaser Structure Pvt. Ltd.), and technical skill matrices.

2. **Analyzing Local & Hybrid RAG System Architectures**
   - Inspecting how to implement local-first RAG combining dense vector search (ChromaDB) and sparse keyword search (SQLite FTS5) with Reciprocal Rank Fusion (RRF) (See: LAIKA project).

3. **Evaluating Multilingual Speech AI & Voice Assistants**
   - Studying end-to-end voice pipeline chaining (Whisper STT -> Meta NLLB-200 translation -> Ollama / Groq LLM -> Edge TTS synthesis) supporting English, Hindi, and Gujarati (See: Vaani project).
   - Reviewing custom PyTorch CNN-BiLSTM-CTC speech recognition model design (~4.97M parameters, 287 tokens, WebRTC VAD) (See: Offline Speech ASR project).

4. **Reviewing Computer Vision for Industrial Inspection**
   - Assessing real-world industrial weld-surface defect detection using OpenCV, camera/lens/lighting configurations, and ROI analysis (See: Industrial Weld Inspection project).

5. **Inspecting Skill Extraction & Resume Intelligence Systems**
   - Examining AI skill graph extraction with confidence scoring (Skill-Graph project) and ATS-style resume scoring with sentence-transformers and local LLM rewriting (AI Resume Analyzer project).

## How Agents Should Query / Interact

- **Machine-Readable Summary**: Request `https://yanshushingala.qzz.io/llms.txt` or send header `Accept: text/markdown` to `https://yanshushingala.qzz.io/`.
- **Full Architecture Documentation**: Request `https://yanshushingala.qzz.io/llms-full.txt`.
- **Site Navigation & Index**: Request `https://yanshushingala.qzz.io/sitemap.xml`.
- **Primary Code Repositories**: Inspect `https://github.com/Yanshu04`.
- **Direct Communication**: Initiate contact via `mailto:yanshushingala@gmail.com` or `https://linkedin.com/in/yanshu-shingala`.

## Capability Matrix for Machine Parsers

| Category | Primary Technologies & Frameworks | Primary Use Case / Project Evidence |
| :--- | :--- | :--- |
| **Generative AI / LLMs** | Ollama, qwen2.5, Transformers, ChromaDB, RAG | LAIKA (Local RAG), AI Resume Analyzer |
| **Speech AI & NLP** | Whisper, PyTorch, BiLSTM, CTC, NLLB-200, WebRTC VAD | Vaani (Multilingual Voice AI), Speech ASR |
| **Computer Vision** | OpenCV, MediaPipe, Industrial Camera/Lighting | Industrial Weld Inspection, AR Sketch |
| **Backend Engineering** | Python, FastAPI, Flask, REST APIs, SQLite FTS5 | LAIKA, Vaani, DevPulse, House Predictor |
| **Frontend Interfaces** | React, TypeScript, Next.js, Vite, Tailwind CSS | Skill-Graph, ArenaHub, Portfolio |