AI / ML Engineer · Researcher · Rutgers Honors

I build intelligent systems that help experts make better decisions.

I'm Sonakshi — an AI/ML engineer and researcher in CS & Data Science. I ship working systems across medical AI, energy, and agentic tools, and I study the questions behind them. I learn fast, work well solo or on a team, and care more about augmenting expert judgment than replacing it.

See how I build → View projects
Status
Building
Agentic systems with a human at the center
Reading
On the calibration of expert trust
Ship

I take systems from question to working prototype, then iterate on real feedback.

Learn fast

New domain, new stack, new lab — I get productive quickly and ask the right questions.

Work with people

I mentor, collaborate, and communicate clearly — solo when needed, better on a team.

Go deep

Comfortable in the weeds of ML, evaluation, and the math — and in explaining it plainly.

Current focus

Three threads at once
Thread 01 · Trust

Agents that show their work

Tool-using systems you can audit step by step, not black boxes you have to take on faith.

Thread 02 · Time

Forecasting that owns its uncertainty

Energy and time-series models that arrive as a calibrated decision, not just a number.

Thread 03 · Care

Medical AI a clinician can interrogate

Explanations that work as a clinical instrument — auditable, grounded, and honest about doubt.

Featured projects

All projects →
001

IntExR

Interpretable, evidence-grounded reasoning for clinical decision support.

Research
002

Energy System Research

Calibrated probabilistic load forecasting with synthetic augmentation.

Research
003

Market Sentiment Prediction

A human-in-the-loop NLP pipeline turning noisy text into decision-ready signals.

Engineering

Featured research

All research →
Area 01 · Healthcare

Explainable Medical AI

Systems that surface auditable, evidence-grounded reasoning a clinician can interrogate before they act.

Explainability Retrieval Human-AI
Area 02 · Energy

Energy Systems & Forecasting

Probabilistic forecasting and synthetic time-series for grid operators making consequential, time-pressured calls.

Time-Series Uncertainty Synthetic Data
The question
I'm sitting with

When an expert and a model disagree, who should win — and how do we design the moment that decides?

Read where this is going →

A quick timeline

2022Rutgers Honors — CS + Data Science, Stats minor
2023Medical & Energy Systems research begins
2024Cornell Break Through Tech AI Fellow · IEEE NLP Fellow
2025Honors Peer Mentor · agentic systems work
Read the full story →

Recent writing

Mar 2025 · Essay
On the cost of an explanation
Jan 2025 · Note
Synthetic data is a modeling assumption, not a free lunch
Nov 2024 · Note
Calibration is a UX problem
All writing →

Let's build something worth trusting.

Open to research collaborations, internships, and conversations about human-centered AI.

Email me ↗ How I build