H
CV

Ahmedabad --:-- IST

01Extract

HarshPatel

The web is noise. I extract the signal.

AI/ML Engineer · LLM Integration · RAG · Automation

10+ LLM APIs, RAG pipelines, agents and local models — powerful AI systems at zero or minimal cost.

years shipping
07+
engineers led
30+
tools mastered
50+
LLM APIs
10+
sites, one pipeline
1,500
Harsh S Patel, smiling, in a light jacket
fig. 01touch & hold to peek

// the human behind the pipelines

Hi, I'm Harsh. I build systems that read the web — and lead the teams that build them.

An AI/ML Engineer and LLM Integration specialist with deep hands-on experience across the modern AI stack. I have integrated 10+ LLM APIs — Claude, ChatGPT, Gemini, Grok, DeepSeek, Perplexity, Mistral, LLaMA, Qwen — and deployed local models via Ollama using free hosting alternatives to save costs.

Expert in RAG pipelines, LangChain, LangGraph, AI Agents, MCP, and vector databases (Chroma, Pinecone, Qdrant, Weaviate, FAISS, Milvus). Built production AI automation systems handling real-world workflows end-to-end without any paid third-party services.

Also skilled in PyTorch, TensorFlow, OpenCV, HuggingFace Transformers, MLflow, Streamlit — covering the full ML lifecycle from experimentation to deployment. My core strength: building powerful AI systems at zero or minimal cost.

Based in
Ahmedabad, Gujarat
Local time
--:-- IST
Shipping since
2018
Coordinates
23.02°N 72.57°E
02Transform

Five transformations. From junior developer to leading 30+.

Starting as a junior PHP developer in 2018, I grew through senior engineering and module leadership — and now lead multiple technical teams.

stage_00 · source()

Education

  • 2014 — 2018B.E. Computer EngineeringLDRP-ITR, Gandhinagar · CGPA 8.5 / 10
  • 2014HSC ScienceGHSEB · 80%
  • 2012SSCGSEB · 89%

stage_01 · ingest()

Junior Developer

First role

Dynamic Web Solutions

Aug 2018 — Aug 2019 · 1 year

Built dynamic websites for multiple clients using PHP MVC architecture and Zend Framework. Delivered full-stack solutions with custom business logic and database design.

  • PHP
  • Zend
  • MVC
  • jQuery
  • MySQL
  • HTML/CSS

stage_02 · deepen()

Senior Developer

Senior dev

Large-scale e-commerce

Aug 2019 — Aug 2022 · 3 years

Core engineer on a major e-commerce platform. Deep expertise in Magento internals, performance optimisation, Solr/Elasticsearch search tuning, and containerised deployments.

  • Magento
  • PHP
  • MySQL
  • Solr
  • Docker
  • Git
  • Postman

stage_03 · orchestrate()

Group / Module Leader

Module lead

E-commerce platform

Aug 2022 — Aug 2024 · 2 years · Team of 12

Led a team of 10–12 developers on a high-traffic e-commerce platform. Owned critical modules: pricing strategy, shipping & tax logic, Google Ads, and competitor monitoring. Architected complex business logic driving direct revenue impact. Mentored mid-level developers and set engineering standards.

  • Magento
  • PHP
  • MySQL
  • GraphQL
  • Elasticsearch
  • Kafka
  • Docker
  • CI/CD
  • JIRA

stage_04 · infer()

Technical Team Lead & Tech Expert

Current

Company-wide AI/ML & automation

Aug 2024 — Present · Now · 20+ members

Leading company-wide AI/ML adoption and automation strategy. Built multiple end-to-end automation systems — real-time monitoring, LLM-driven processing, smart scheduling — all without paid tools. Integrated 10+ LLM APIs including local models via Ollama. Built enterprise-scale scraping pipelines handling CAPTCHAs and anti-bot systems. Managing 20+ members, providing internship training.

  • LangChain
  • RAG
  • Ollama
  • MCP
  • Scrapy
  • Playwright
  • n8n
  • Airflow
  • FastAPI
  • Vector DB
  • AWS
  • MLflow
  • HuggingFace
03Load

case_01 · 1500_sites.etl

Case file: 1,500 websites, one schema

0000 / 1500 sites extracted

  • ? CAPTCHAs
  • # Login walls
  • { JS rendering
  • % Anti-scraping

↓ 1 unified schema

01 · The brief

1,500 websites. Each one fighting back.

A client building a domain-specific LLM needed structured data from nearly 1,500 websites — every one with its own obstacles: CAPTCHAs, login walls, JavaScript rendering and anti-scraping measures.

02 · The framework

One framework. A configurable adapter per site.

My team of 5–6 designed a standardized extraction framework with configurable adapters per site — so every new site was configuration, not a rewrite.

03 · The run

Queue. Retry. Extract. Then normalize everything.

A queue-based workflow with automatic retries kept the crawl moving, while a centralized ETL layer normalized every source into one unified schema.

04 · The result

3× faster. All 1,500 sites in 2–3 months.

We finished running 3× faster than the client's own in-house team working in parallel. The output was clean and ready for LLM training with minimal rework.

More case files

  • case_02

    Price monitoring at e-commerce scale

    Ongoing price monitoring systems for e-commerce clients — millions of product-match records with real-time alerting pipelines.

    • Millions of records
    • Product matching
    • Real-time alerts
  • case_03

    AI automation systems, zero paid tools

    End-to-end automation systems — real-time monitoring, LLM-driven processing and smart scheduling — built without a single paid tool.

    • Real-time monitoring
    • LLM processing
    • Smart scheduling
  • case_04

    10+ LLM APIs, plus local models

    Integrated Claude, ChatGPT, Gemini, Grok, DeepSeek, Perplexity, Mistral, LLaMA and Qwen — and deployed local models via Ollama on free hosting alternatives to save costs.

    • 10+ APIs
    • Local models
    • Ollama
  • case_05

    Revenue-critical e-commerce modules

    On a high-traffic Magento platform: pricing strategy, shipping & tax logic, payment gateway integration, Google Ads management, competitor monitoring and Solr/Elasticsearch search tuning.

    • Magento
    • Pricing · shipping · tax
    • Search tuning
04Embed

78 tools, embedded. Related tools cluster together.

All skills by category

Languages

  • Python
  • PHP
  • JavaScript
  • HTML/CSS
  • jQuery
  • GraphQL

Frameworks

  • FastAPI
  • Django
  • Flask
  • Magento
  • Zend
  • Odoo

Databases & search

  • MySQL
  • PostgreSQL
  • BigQuery
  • ClickHouse
  • Elasticsearch
  • Solr

Orchestration & agents

  • LangChain
  • LangGraph
  • LangSmith
  • n8n
  • Airflow
  • Celery
  • AI Agents
  • MCP

ML libraries

  • PyTorch
  • TensorFlow
  • Keras
  • Scikit-learn
  • Transformers
  • HuggingFace
  • OpenCV
  • Pandas
  • NumPy
  • MLflow
  • Streamlit
  • Gradio

LLMs & prompting

  • Claude
  • ChatGPT
  • Gemini
  • Grok
  • DeepSeek
  • Perplexity
  • Ollama
  • Mistral
  • LLaMA
  • Qwen
  • Prompt Engineering

RAG & vector DBs

  • RAG
  • Embeddings
  • Chroma
  • Pinecone
  • Weaviate
  • Qdrant
  • FAISS
  • Milvus
  • Amazon S3 Vectors

Scraping & automation

  • Scrapy
  • Playwright
  • Selenium
  • Puppeteer
  • BeautifulSoup
  • httpx
  • aiohttp
  • Anti-CAPTCHA
  • Proxy Rotation
  • Redis

DevOps & tools

  • Docker
  • AWS
  • Git
  • GitHub Actions
  • Jenkins
  • CI/CD
  • Kafka
  • JIRA
  • Notion
  • Postman

Scroll to tour the clusters, or point at any label. ◆ marks what this lens cares about.

05Retrieve

Ask the index. It answers in milliseconds, for free.

A tiny retrieval system over everything on this page — the same idea behind the RAG pipelines I build, minus the API bill.

The answer, its sources and their scores show up here. Matching skills light up in the field behind.

Runs in your browser · BM25 over 41 hand-written notes · no LLM · $0 per query — or ask the human: harshpatel344930@gmail.com

06Optimize

“Every paid tool has a free alternative

— and I always find it.”

— core philosophy
Departures · paid stackMonthly bill: $$$$
  • LLM inferencePAID API CALLS replaced by LOCAL MODELS / OLLAMA
  • Vector searchMANAGED VECTOR DB replaced by CHROMA / FAISS / QDRANT
  • WorkflowsAUTOMATION SAAS replaced by N8N / AIRFLOW
  • Data extractionSCRAPING APIS replaced by SCRAPY / PLAYWRIGHT
  • Job queuesMANAGED QUEUES replaced by CELERY / REDIS
  • Model hostingPAID GPU HOSTING replaced by FREE ALTERNATIVES

Systems architected to save thousands every month across data, LLM and SaaS — without giving up production quality.

07Orchestrate

Lead by example. Hands-on technically, strong on delivery.

2018 — 2022

Individual contributor. Four years hands-on.

PHP MVC and Zend for clients, then Magento internals, performance and search tuning on a large-scale platform — before leading anyone.

2022

A team of 10–12. Revenue-critical modules.

As module leader on a high-traffic e-commerce platform: pricing strategy, shipping & tax logic, Google Ads and competitor monitoring. Mentored mid-level developers, set engineering standards.

2024

A 20+ member team. Automation & scraping.

Defining processes, setting quality standards and mentoring engineers at every level — while leading company-wide AI/ML adoption.

Always

Interns, trained. Freshers, made industry-ready.

Internship training in PHP, Python, AI/ML and web scraping — bridging academic knowledge and industry-ready skills.

engineers led in total
30+
Magento e-commerce team
10–12
automation & scraping team
20+
trained in PHP · Python · AI/ML
Interns

Key strengths

  • { }

    Web scraping at scale

    Dynamic sites, CAPTCHAs, anti-bot systems at production speed. Built & trained a 20+ member scraping team.

  • ~>

    LLM & RAG engineering

    Integrated 10+ LLM APIs and local models via Ollama. Expert in RAG, vector search, embeddings and agents.

  • </>

    Full stack engineering

    PHP, Python, Magento, FastAPI, Django. Complex business logic, e-commerce systems, and scalable APIs.

  • o-o

    Team leadership

    Led 30+ member teams across dev, automation and scraping. Mentored freshers, set standards, drove delivery.

  • <->

    AI automation systems

    End-to-end automation — real-time monitoring, LLM-driven processing, smart scheduling — zero paid tools.

  • ₹0

    Free over paid

    Every paid tool has a free alternative. Architect systems saving thousands monthly across data, LLM, and SaaS.

08Respond

Send a signal. I reply within 24 hours.

A freelance project, a full-time opportunity or a quick consultation — scraping, AI/ML, automation, backend or team training.

POST /harsh · content-type: human

harshpatel344930@gmail.com
Write to me →

// a good first message has

  1. 01What you're building, extracting or automating
  2. 02A rough timeline
  3. 03The best way to reach you

Direct channels