Chirag Pansuriya

AI & Automation

Experience with models inside real products.

I have shipped LLM features with the OpenAI API and Azure AI Services: integration, workflow automation, prompt engineering, and AI evaluation. Open to full-time, contract, and freelance work in this area.

  1. 01OpenAI API

    I have used the OpenAI API inside a live support product, not as a demo chat. On Deskly AI the API handles FAQ answers, booking help, ticket summaries, and suggested-reply drafts. The call sits in the same flow the staff already use.

  2. 02Azure AI Services

    Deskly AI also runs on Azure AI Services for the same support jobs: chat, summaries, and draft replies. I have worked with both OpenAI and Azure on one product, so the model provider can change without throwing away the workflow around it.

  3. 03LLM integration

    Integration means the model is part of the product that already has users. Deskly AI puts LLM chat and drafts into support for small service businesses. Jaagruk Bharat adds short summaries on top of civic feeds that were already on the App Store and Google Play. The screen, the data, and the model stay in one path.

  4. 04AI-assisted workflow automation

    On Deskly AI I designed prompt templates and automated workflows that produce a draft and then stop. Staff review and edit the AI output before it is sent. The automation saves the first pass. A person still owns the message that goes out.

  5. 05Prompt engineering

    I write the prompt for the job the product actually has. Deskly AI uses prompt templates for FAQ chat, booking help, ticket summaries, and suggested replies. On Jaagruk Bharat the prompts have to stay faithful to the source update, so a summary does not add claims that were not in the feed.

  6. 06AI evaluation

    I read the output before it ships. I check accuracy, relevance, consistency, and whether the model followed the instruction. On Jaagruk Bharat, summary runs were compared before one went live. On Deskly AI, evaluation is also the staff review step: a wrong draft does not get sent.

Where this shipped

  • Deskly AI

    Current. OpenAI API and Azure AI Services for FAQ chat, booking help, ticket summaries, and suggested replies. Prompt templates and workflows with human review before send.

  • Jaagruk Bharat

    App Store and Google Play. LLM summaries on existing feeds, with prompt checks and AI evaluation before a summary went live.

Available for this work on full-time, contract, and freelance terms. The contact page has email and LinkedIn.

Contact