Browse categories
Explore
Fiverr Pro
English
$
USD
I will build production llm features for your product
Edgardo P
About this gig
I build production AI features you can ship, not ChatGPT demos.
I integrate OpenAI and Claude into real products: RAG over your documents, tool-use agents, streaming chat, structured JSON, evals, and guardrails. I work in Python, TypeScript, LangChain, FastAPI, and Node.
What you get:
- An AI feature that fits your existing app or API
- Retrieval, tool calls, or chat flows with error handling
- Eval cases so a model change does not silently break you
- Run steps, env notes, and how to extend it
Good fit if you need:
- Chat, support, or review-reply features in your product
- RAG over PDFs, docs, or a knowledge base
- Agents that call your APIs safely
- An LLM prototype made production-ready
Not a fit for training a model from scratch, or for jailbreak work.
Send the use case, the model, and your deadline. I will reply with a plan and the right package.
Related work: TryReplify, an LLM reply generator with tone, language options, and quota-aware usage.
Get to know Edgardo P
Edgardo P
Software Engineer
- FromPhilippines
- Member sinceMay 2026
- Avg. response time1 hour
Languages
English
I'm an engineer who ships LLM-powered product features end to end, with 8+ years of full-stack delivery behind me. I design retrieval and agent-style workflows, integrate OpenAI and Anthropic (and similar) APIs, and care about the whole production story: latency, cost, safety, and observability-not only the demo.
I build the surrounding software the same way: React and TypeScript on the front, .NET and Python (FastAPI, Django) on the back, and AWS/Firebase in the cloud. That background matters when AI has to sit inside real auth, billing, data policies, and scale.
FAQ
Which models do you build with?
OpenAI (ChatGPT) and Anthropic (Claude) first. If you use another chat API, send the model name and I will confirm before you order.
Will you log into my OpenAI or Anthropic account?
No. You keep the keys. I write the integration and, if needed, a small test script you can run. If we must test inside your project, use a restricted key and share it only through Fiverr.
Do you also write prompts only, with no code?
This gig is AI development: code, retrieval, and wiring into your product. Prompt-only work is not offered.
Can you take an existing prototype and make it production-ready?
Yes. Paste the repo or current flow, 3 failing examples, and the outcome you want. Standard is usually the right package.
What is an eval set?
A small list of inputs and expected outputs I use to check the feature. It is how we know a revision did not break the cases that already worked.
Do you train custom models or write jailbreaks?
No. I will not train a model from scratch, and I will not write code or prompts meant to bypass safety rules, scrape private data, or impersonate people.
