Mehul D.
AI Engineer
I build AI agents, RAG chatbots and LLM integrations, and I wire them into the systems a business already runs on. On the backend I work mostly in Python with FastAPI, and in Laravel, with React on the front end. For the AI side that means the OpenAI, Claude and Gemini APIs, LangChain and LangGraph for agent logic, vector databases for retrieval, and n8n when a workflow tool gets the job done faster than custom code. Getting a first demo to work is usually quick. Where I spend my time is everything after that: answers that point back to the right document, agents that hand off to a person when they aren't sure, tool calls that fail cleanly, and token costs that stay predictable as usage grows. Projects I can walk you through: » A multi agent platform that sends each request to the agent and model best suited for it, across OpenAI, Claude and Gemini. I built the role-based agent orchestration and the RAG layer that lets users work with their own documents. » An AI agent for agencies. It keeps account records current without anyone typing them in, sits in on client calls, drafts proposals against the agency's own pricing and scoping rules, and answers questions in Slack and Teams. » Voice AI agents for customer support, sales and recruitment calls, with WhatsApp, chat and email on the same platform. I set up the backend architecture, did the AI implementation and integrations, and built the backend flow that carries each conversation through to its outcome. » Insurance An AI operations assistant for a 30-year-old insurance consultancy. Each agency gets its own assistant trained only on its SOPs, and it won't answer outside that scope. It also drafts audit guidelines for each agency, which admins approve before release. Audit handling time dropped 45%. » Ecommerce AI fraud detection on bulk orders, product description generation, ad analysis and image pipelines for a large US ecommerce platform, all running in production at high order volume. » Healthcare Patient intake and therapy adherence systems for a NASDAQ listed healthcare company serving millions of patients in 40+ US states. Intake replaced manual fax triage and re-keying. » Email automation (n8n) A workflow that sorts incoming mail into support, sales and job applications, drafts a reply for the right branch, logs everything to Airtable and posts summaries to Slack. ✦ Stack: Python · FastAPI · Laravel · PHP · React · Node.js · LangChain · LangGraph · OpenAI · Claude · Gemini · Copilot Studio · Hugging Face · Pinecone · FAISS · Weaviate · PostgreSQL · MySQL · Redis · Docker · AWS · GCP · n8n · Make · Zapier · Twilio · ElevenLabs If you send the job post along with a sample of your documents or data, I'll tell you how I'd build it and where I expect it to break.
Earnings and projects
Last 6 months ÷ 6
Last 6 months: $5.6k
Billed in all: $586,855$587k
$1.9k in the last 6 months
Across 193 contracts
Earnings, last 30 days
Skill rankings: Top 8% in CodeIgniter
Ranked by MRR among everyone on Upwork with the skill.
- #31OCR Algorithm183 freelancers · 🇺🇸 #3Top 17%🇺🇸#3 of 9
- #55Shipping & Order Fulfillment Software244 freelancers · 🇺🇸 #4Top 23%🇺🇸#4 of 26
- #135LLM Prompt799 freelancers · 🇺🇸 #10Top 17%🇺🇸#10 of 72
- #268CodeIgniter3,448 freelancers · 🇺🇸 #11Top 8%🇺🇸#11 of 65
- #367ASP.NET2,819 freelancers · 🇺🇸 #48Top 13%🇺🇸#48 of 281
- #980Vue.js8,267 freelancers · 🇺🇸 #61Top 12%🇺🇸#61 of 495
- #1,373Laravel12,549 freelancers · 🇺🇸 #89Top 11%🇺🇸#89 of 523
- #1,445React Native10,486 freelancers · 🇺🇸 #144Top 14%🇺🇸#144 of 697