Simra Shahid

À propos

I am a Researcher II in Microsoft Research’s PROSE group. My research focuses on getting people the information they need, in a form they can act on, when it matters most. I study how user’s domain, culture, and expertise shape what counts as “useful,” then design adaptive systems that bridge those gaps. I’m currently investigating how AI can act as an “attention-debiaser” to uncover missed opportunities in data visualisation and scientific discovery.

Before joining Microsoft, I was a Research Scientist on Adobe’s Media & Data Science team, contributing to projects all the way from early research to product launch. My work on document-recommendation, conversational-assistant, and personalisation pipelines now supports various products in Adobe including Adobe Customer Journey Analytics, Experience Platform, Marketo, and the Community Platform.

• Conversational assistants: We reduced routine load for sales and marketing teams by automating answering common questions, booking calls, and routing harder issues or edge-cases to human agents. The stack combined retrieval-augmented Q&A, chat summarisation with guardrails, query standardisation, and improving dialogue-flow design.
• Content strategy: We connected granular attributes of content (text and images) and connected to audience signals to rate content assets, highlight competitive gaps, and guide marketers toward higher-impact content choices (Catchy Content et Content Tensor).

Recent Research Projects:

• Scideator et Idea Novelty Checker: an LLM-powered workflow that helps researchers combine concepts across disciplines and check novelty against the literature.
• “How to Take Back Control of Your Data“:  techniques to limit large language models from memorizing sensitive customer data without access to model weights. [Blog] [Paper]
• CoPrompter: User-Centric Evaluation of LLM Instruction Alignment for Improved Prompt Engineering. [Blog] [Paper]
  A tool to help prompt engineers compare their LLM responses using a user-editable checklist and automatically flags where LLM responses fall short, giving prompt engineers a fast, systematic way to refine alignment.

Selected Recognition:

• Outstanding Young Engineers Award. One of seven in Adobe to receive this award.
• Innovation Ambassador. National-Level Award, given to five entrepreneurs in India by Government of India and the
  Deputy High-Commission of Canada-India for advancing “Blue is Your Clue” proposal to an international stage in Canada.
• Recognition from Indian Prime Minister et Singapore Health Minister (Link).