You will get a custom AI agent that automates repetitive work
Project details
If your team loses hours every week to the same manual task (moving data between tools, tagging tickets, pulling info out of documents), that's exactly what an AI agent should do instead of a person.
Take one repetitive workflow off your plate and have it run on its own, so your people get that time back for work that needs them.
I build custom AI agents that string together the steps you do by hand. That means a few tool integrations, an LLM step where judgment is needed, and guardrails so it doesn't go off the rails. I've built an LLM-powered pipeline for an enterprise client that auto-generates test cases and uses models to find and debug failures, so production-grade agent work is my home turf.
How I'd approach yours: I map the process first, work out where an LLM genuinely helps and where it shouldn't touch anything, then build and test against your real examples before it goes near live work.
I'm newer on Upwork and putting everything into getting these first jobs right. We can start with a small milestone you release only once it works. More of my work is in my portfolio.
Tell me the task eating your team's time and I'll tell you honestly if an agent can handle it.
Take one repetitive workflow off your plate and have it run on its own, so your people get that time back for work that needs them.
I build custom AI agents that string together the steps you do by hand. That means a few tool integrations, an LLM step where judgment is needed, and guardrails so it doesn't go off the rails. I've built an LLM-powered pipeline for an enterprise client that auto-generates test cases and uses models to find and debug failures, so production-grade agent work is my home turf.
How I'd approach yours: I map the process first, work out where an LLM genuinely helps and where it shouldn't touch anything, then build and test against your real examples before it goes near live work.
I'm newer on Upwork and putting everything into getting these first jobs right. We can start with a small milestone you release only once it works. More of my work is in my portfolio.
Tell me the task eating your team's time and I'll tell you honestly if an agent can handle it.
AI Algorithms
Large Language Model, Transformer ModelAI Applications
AI Content Creation, AIOps, Conversational AI, Natural Language UnderstandingAI Models
ChatGPT, GPT-4, LLaMAWhat's included
| Service Tiers |
Starter
$120
|
Standard
$400
|
Advanced
$1,100
|
|---|---|---|---|
| Delivery Time | 5 days | 10 days | 18 days |
Number of Revisions | 2 | 3 | 4 |
AI Model Integration | - | - | - |
Batch Normalization | - | - | - |
Database Integration | - | - | - |
Detailed Code Comments | - | - | - |
Image Upscaling | - | - | - |
MLOps | - | - | - |
Model Deployment | - | - | - |
Model Documentation | - | - | - |
Model Monitoring | - | - | - |
Model Testing & Optimization | - | - | - |
Model Tuning | - | - | - |
Natural Language Processing | - | - | - |
NLP Tokenization | - | - | - |
Pre-Training | - | - | - |
Prompt Engineering | - | - | - |
Setup File | - | - | - |
Source Code | - | - | - |
Frequently asked questions
About Somesh
AI Automation Engineer | RAG & LLM Systems | Python | Local AI | CCF-A
Hyderabad, India - 4:05 am local time
I built an LLM code testing system for Thomson Reuters ($35B+), and I rebuilt a code intelligence platform that now runs at five tech companies.
The hard part of AI was never the demo. It's building something that still works once real users get their hands on it. Most of the time the problem isn't the model, it's what gets pulled in before it answers, and that's the part you can actually fix. That's the work I'm good at.
I'm an AI and LLM engineer, and I also work full stack in Python (FastAPI) and Next.js/React, so I can take an AI feature from a rough idea all the way to a shipped product. I'm Anthropic Claude Certified, and there's real work of mine in the portfolio below.
A few things I've built:
→ Thomson Reuters ($35B+): an LLM code testing system (Dockerized Python + Next.js) that reads GitHub repos, writes its own test cases, and uses LLMs to find and fix failures. I got pulled onto the client team in my first week.
→ A code intelligence platform I rebuilt from an earlier version. It turns any codebase into a graph you can query, then writes documentation and architecture diagrams off it. It now runs at five tech companies.
→ A meeting assistant (Whisper, Pyannote, local SLMs) that transcribes and summarizes entirely on the user's own machine, so no audio ever leaves it. Running inside EPAM Systems today.
→ A live dictation tool that turns speech into clean text in under a second, fully offline.
What I can build for you:
→ RAG systems and chatbots over your own docs, PDFs, or knowledge base (LangChain, vector DBs, OpenAI/Claude), with sources you can check so it won't invent answers
→ AI agents and workflow automation (Make dot com, Zapier, n8n, custom API pipelines)
→ LLM features inside your app, plus the FastAPI and Next.js backend and frontend to run them
→ Evaluation and testing for your LLM, so the wrong answers get caught before your users find them
→ AI that runs on your own hardware, for when your data can't touch the cloud (legal, finance, health)
How I work: I nail down scope early, keep you in the loop as I go, and hand you clean, documented code you can maintain long after we're done. I build things meant to survive real traffic, and I take security seriously, having found and reported real auth vulnerabilities the right way.
If you're building an AI feature, a chatbot, or an automation, send me a short brief and I'll tell you straight how I'd tackle it. I usually reply within a few hours.
Steps for completing your project
After purchasing the project, send requirements so Somesh can start the project.
Delivery time starts when Somesh receives requirements from you.
Somesh works on your project following the steps below.
Revisions may occur after the delivery date.
Show me the manual process
You walk me through the task you keep doing by hand, the tools involved, and what a good result looks like. Real examples help me far more than a written spec.
I map the workflow
I sketch out the steps, where an LLM genuinely helps, and where it shouldn't touch anything. You see the plan before I build, so there are no surprises later.