Open to opportunities

Akansha Sharma

@akanshasharma3

Message

I build multimodal GenAI systems that make complex information accessible and actionable.

India

Message

What I'm looking for

I'm looking to build accessible, real-world AI solutions, especially multimodal systems using LLMs, vision-language models, OCR, and RAG. I value opportunities to research, fine-tune, deploy, and evaluate scalable AI products.

At the National Institute of Technology, Durgapur, I built multimodal AI systems for accessibility, including a LangChain, Hugging Face, and PaddleOCR pipeline that describes mathematical expressions for visually impaired users in real time. I also developed a Flask-based Chrome extension that improved accessibility efficiency by 40% and mentored three junior members, increasing lab productivity by 25%.

I've built healthcare, document intelligence, audio analysis, and mathematical reasoning systems using LLMs, vision-language models, OCR, RAG, and structured evaluation. My work includes MediLens for prescription understanding and drug interaction analysis, FormGen-AI for extracting structured data from PDFs and images, and published research on factual reasoning in language models and state-aware prescription risk flagging.

Experience

Work history, roles, and key accomplishments

AI & STEM Accessibility Research Intern

National Institute of Technology, Durgapur

Feb 2024 - Apr 2026 (2 years 2 months)

Built a multimodal AI pipeline using LangChain, Hugging Face, and PaddleOCR to describe mathematical expressions for visually impaired users in real time. Developed a Flask-based Chrome extension improving accessibility efficiency by 40% and mentored 3 junior members, increasing lab productivity by 25%.

Lang Chain Hugging Face Paddle OCR Flask Multimodal AI Accessibility Mentor