Hi, I'm Pranav.

AI Automation Engineer

Ahmedabad, India

Open to Remote & Relocation

I build AI systems that eliminate the work nobody wants to do.

What I've eliminated

01Manual lead sorting
028-hour MIS reports
03Customer reply queue
How this assistant works

Built for accurate answers.

01

Intent-aware routing

GPT-4.1 nano reads your question with recent conversation context, resolves typos and follow-ups, then returns a structured intent and search query.

02

Filtered retrieval

The query is embedded with text-embedding-3-small and searched in Qdrant. Category filters, document priority, recency, and similarity reranking select resume-backed evidence.

03

Grounded response

GPT-4o mini streams an answer from the selected context and adds source citations. If the portfolio has no supporting information, it says so instead of guessing.

Ask Pranav's workAnswers grounded in real projects and experience

Pick a question below, or type your own.

Try one of these

Shift+Enter for new line · powered by a custom RAG pipeline

Selected work

Projects that actually shipped

Production systems serving real businesses. Open any project for the complete technical and business view.

Open a project to explore its decisions, architecture, learnings, and outcome.
Career

Where I've worked

Junior Software Developer — AI & Automation

Current

Tech Grind Pvt. Ltd.·AI & software product company · Ahmedabad, On-site

Jul 2025 – Present

Own the full AI and automation workstream — from event-driven lead intake across WhatsApp and CRM to self-hosted RAG pipelines serving multilingual customers. Every system I ship here has a measurable time-saved or effort-reduced outcome.

  • Replaced manual lead sorting with an event-driven WhatsApp-to-CRM workflow using LLM intent classification
  • Automated MIS reporting with validation and exception handling — saving 150–200 hours per month
  • Built a multi-agent WhatsApp platform with role-based conversation routing for support and sales
  • Deployed a self-hosted RAG pipeline (Qwen2.5 via Ollama) keeping multilingual response times under 5 seconds
  • Led a product-research automation effort that cut manual research effort by ~85%
n8nLangChainFAISSOllamaNode.jsFastAPIRAGWhatsApp API

Full-Stack SDE Intern

RedSand Technology·Tech company · Dubai, UAE — Remote

Feb 2025 – May 2025

Delivered frontend features for salary analysis and property inspection dashboards, shipped a WhatsApp chatbot that automated 80% of customer support queries, and built a LinkedIn data enrichment pipeline for the sales team.

  • Owned frontend delivery for salary analysis and property inspection dashboards with role-based access control
  • Automated ~80% of customer queries by building a WhatsApp chatbot integrated into existing support workflows
  • Built a data enrichment pipeline that turned raw LinkedIn data into structured company and contact insights
ReactNode.jsWhatsApp APIData EnrichmentDashboard

Web Developer Intern

BasketHunt Pvt. Ltd.·E-commerce startup · Gurugram — Remote

Jul 2023 – Sep 2023

Improved operational efficiency by integrating Zoho and Odoo, and managed user provisioning through Google Admin.

  • Improved operational efficiency by 15% by integrating Zoho and Odoo
  • Managed user provisioning and access control through Google Admin
ZohoOdooGoogle AdminIntegrationWeb Development
Technical skills

The tools I use — and why

Hover a skill to see a one-line rationale. Every choice has a reason.

AI & LLM

Building on top of language models at the systems level — RAG, agents, and production LLM pipelines.

LangChain
Primary orchestration framework for RAG and agent pipelines
RAG Architecture
Vector indexing, chunking, hybrid retrieval, hallucination guardrails
FAISS
Go-to for self-hosted vector search
Ollama
Self-hosted LLM serving — Qwen2.5, Llama, Mistral
Prompt Engineering
Intent classification, structured extraction, grounding
Function / Tool Calling
Conversation Memory
Persistent state across multi-turn agent conversations
LLMOps
Evaluation, observability, model versioning

Workflow Automation

Replacing manual steps with durable, event-driven pipelines.

n8n
Primary tool for complex multi-system workflow orchestration
Cron Jobs
Scheduled pipelines for MIS reporting and batch processing
Webhooks
WhatsApp, CRM, and third-party event triggers
Redis
Job queues, caching, and session state
Bull Queue
Node.js job queue for distributed worker pools
Playwright
Headless browser automation for crawling pipelines

Backend Engineering

The API layer that makes AI systems production-ready.

Node.js
FastAPI
Primary Python API server for AI/retrieval services
Python
Flask
Express.js
REST APIs
Design, auth (JWT), versioning, and integration
JWT Auth

Frontend

Enough to ship polished, interactive products end-to-end.

React.js
Next.js
App Router, Server Components, streaming
Tailwind CSS
Framer Motion

DevOps & Infrastructure

Deployment, containerisation, and keeping things running.

Docker
Compose for local dev; containers for production AI services
Git
GitHub Actions
Linux / SSH
About me

A little more human

Pranav Thakwani — AI Automation Engineer
Get in touch

Got something interesting?

Send me the messy version of the idea or problem. We can figure out the rest from there.