🫀 building trustworthy medical ai · 🔍 studying how models revise their answers · ☁️ turning research into real systems
ml/ai engineer @ rwjms · ai/ml fellow @ microsoft · medical ai & energy systems researcher · cs + data science @ rutgers honors
🫀 int.ex.r — developing clinician-in-the-loop medical ai that supports retrieval, dialogue, and explanation refinement
🏥 health equity readiness (her) — building a human-in-the-loop research platform at rwjms for processing interviews, reviewing ai-generated outputs, protecting participant privacy, and producing research-ready insights
🐙 repository trust signals — working with microsoft to study how developers evaluate the trustworthiness of open-source repositories
⚡ synthetic energy futures — using diffusion models to generate realistic energy-system time-series data
☁️ end-to-end healthcare ai infrastructure — developed and deployed an aws-based workflow using bedrock, lambda, step functions, dynamodb, s3, cognito, comprehend, transcribe, amplify, and quicksight
🤝 human-in-the-loop ai — designed multi-stage review workflows that keep researchers in control of transcription, privacy, qualitative coding, and analytics
🎾 tennis + nlp — explored whether press-conference language can help predict professional tennis upsets
🥗 nourri — co-led the development of a food-scoring framework combining nutrition, processing, and sustainability signals
- trustworthy and explainable ai
- human-ai collaboration
- medical ai
- llm evaluation and revision
- agentic systems
- generative modeling
- applied machine learning
machine learning
python pytorch transformers scikit-learn xgboost langchain
data
pandas numpy pyspark sql
cloud + systems
aws bedrock lambda step functions dynamodb s3 cognito git
languages
java c
when new evidence arrives, can ai change exactly what it should—and nothing else?
💼 linkedin: linkedin.com/in/sharma-sonakshi
📧 email: s.sonakshi2006@gmail.com
🐙 github: github.com/sonak11
🏏 cricket addict
🎾 former nationally ranked junior tennis player
📚 collecting papers faster than i can read them
☕ powered by curiosity and caffeine