DocBot

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for DocBot

DocBot – AI Assistant for Medical Documents ✨ Inspiration

Doctors lose precious time scanning prescriptions, lab reports, and discharge summaries. This slows down patient care and can lead to errors. We wanted to build something that gives doctors instant answers from patient records, so they spend more time treating patients and less time reading files.

📚 What We Learned

How to use NLP + embeddings to turn documents into searchable vectors:

Embedding(d)=f(d)∈R^n

How to apply cosine similarity to find the most relevant medical details:

cos(θ)=A⋅B/∥A∥∥B∥

​

How to combine vector search (FAISS) with an LLM to build a Retrieval-Augmented Generation (RAG) pipeline.

🛠️ How We Built It

Upload & Chunking: Doctors upload prescriptions or reports → text is split into chunks.

Embeddings: Each chunk converted into vectors with all-MiniLM-L6-v2.

Storage: Vectors stored in FAISS for fast retrieval.

Query: Doctor’s question → embedded → compared with stored vectors.

Answer: Relevant chunks sent to LLM (Gemini/GPT) → concise medical response returned.

UI: A Flutter chatbot interface connected via a FastAPI backend.

⚡ Challenges

Handling handwritten prescriptions with OCR.

Managing medical abbreviations like bid, OD, Rx.

Keeping patient data secure and private.

Optimizing latency so doctors get answers in <1 second.

Outcome

DocBot can:

Extract medicines, dosage, allergies from prescriptions.

Summarize lab reports & discharge summaries.

Let doctors ask “What medicines is this patient on?” and get instant, accurate answers.

Tagline: “Smarter records, better care.”

Built With

• ai
• amazon-web-services
• apis
• built-with-languages:-python-?
• cloud
• containerization)
• dart-frameworks:-backend:-fastapi-(python)
• databases
• docker
• embeddings
• faiss
• for
• gemini
• generative
• github
• google
• gpt
• handwritten
• huggingface
• integrations:
• learning
• llm
• machine
• nestjs-(typescript)-frontend:-flutter-(mobile)
• next.js
• nlp:
• ocr
• openai
• postgresql
• prescriptions
• relational
• storage:
• tesseract)
• transformers
• typescript
• typesense
• vector
• web)