Sharyu Deshmukh

Product Data Scientist

My work sits at the intersection of real-world evidence and product. I have 6+ years of experience in healthcare analytics, defining north-star metrics, building studies and pipelines, and shipping AI tools where the model interprets and tested code analyzes.

Explore my work

Work

1
Natural-language question
e.g. how persistent are patients on drug X after first line
2
LLM interpretation
parsed intent and time windows, kept separate from code
3
Cohort specification
editable assumptions with validation checks
4
Deterministic SQL + stats
DuckDB cohort build, persistence estimates
5
Auditable output
results with review flags and provenance

Question → auditable answer, in five steps

Independent project

In evaluation

RWE Insight Copilot

I built an AI-powered copilot that translates natural-language treatment-persistence questions into structured cohort specifications, connecting pharmaceutical analytics needs with reproducible study workflows. An auditable workflow separates LLM interpretation from deterministic cohort and statistical code, with editable assumptions, instrumented validation checks, and analyst review flags.

Editable assumptions, review flags
Analysts can adjust time windows and inclusion rules, and see validation checks before trusting any output.
Evaluated on 30 hand-checked questions
Specification accuracy, clarification behavior, and summary faithfulness measured before any release.

Python
DuckDB
Claude API
Streamlit

1
Garmin data export
daily workouts, sleep, and HRV
2
Metric definitions + data-quality rules
readiness tracked per muscle group
3
Recovery model
tomorrow's workout with the three numbers behind it
4
Built-in check
did the advice actually help? tracked like any product KPI

From watch data to tomorrow's workout

Data product · solo

Live

NextRep

I built a daily “what should I train next?” recommender on my own Garmin data — metric definitions, data-quality rules, and a built-in check on whether the advice drives retention. I built it around Garmin closing its developer API, at zero running cost.

Daily next-workout card
Names tomorrow's workout, shows the three numbers behind it, and offers a low-energy alternative to keep daily engagement high.
Recovery by muscle group
Readiness per muscle after every session, traced back to the workouts that caused the fatigue.

Learn more

Python
SQL
Postgres RLS
GitHub Actions
React

Explore project: live demo

1
Healthcare claims data
5TB+ integrated from three vendors
2
AWS pipelines
scalable SQL transformations on AWS
3
Self-service apps
patient/physician journeys, health-economics models
4
Researcher answers
adoption and engagement without an analyst in the loop

Data to self-service, in four layers

Anlitiks, Inc.

In production

RapidAnalyzer

I own product analytics for RapidAnalyzer, a real-world-evidence analytics SaaS platform, and serve as the product SME between customers, Product, and Engineering. I led the agile build of a health-economics model framework and a patient/physician journey app so researchers answer questions themselves.

Researchers answer questions themselves
Built patient/physician journey and health-economics apps so studies move without an analyst in the loop.
Product analytics from Beta to Production
Defined the KPIs for usage, engagement, and feature adoption and built the self-service dashboards behind monthly product deep dives.
5TB+ claims data, reliably integrated
Architected AWS pipelines across three data vendors to power large-scale healthcare analytics.

AWS
SQL
R / RShiny
Tableau
A/B testing

Experience

Anlitiks, Inc.

Jul 2022 – Present

Data platform and services company delivering real-world evidence (RWE) analytics for pharma and life-sciences clients

I own product analytics for RapidAnalyzer and lead real-world-evidence study delivery for pharma clients.

Anlitiks, Inc.

Apr 2020 – Jun 2022

I built the analytics that turned recurring research requests into self-service tools.

State Street Corporation

Jan 2019 – Aug 2019

I benchmarked the bank's credit risk models and automated a three-decade sovereign rates workflow.

RWE Insight Copilot

2026

Independent project

I built an AI-powered RWE copilot to make first-look persistence analyses defensible in minutes.

About

I'm a product data scientist with 6+ years of experience, currently at Anlitiks, a healthcare data startup. My work sits at the intersection of business, research, and engineering: I define the metrics that matter, translate complex business needs into data-driven solutions, lead end-to-end studies on terabytes of healthcare claims data, and build the tools and AI products that let researchers answer questions themselves.

I live in the Pacific Northwest with my vizsla, Scotch. Outside of work, I love spending time outdoors, practicing yoga, and experimenting in the kitchen.

Publications

“Disease Burden, Healthcare Utilization & Costs of Primary Sclerosing Cholangitis in the US” — Liver International Communications.

Education

M.S. in Data Science · Northeastern University, Boston, MA · Dec 2019
B.E. in Computer Science · RCOEM College, Nagpur, India · Jul 2016

Certifications

AI Product Manager Specialization, Duke University
Databricks Generative AI Fundamentals
A/B Testing, Google (Udacity)
Product Analytics Micro-Certification, Product School

Statistics & causal inference

Propensity-score matching
Survival analysis
Cox models
Hypothesis testing
A/B testing

Data & engineering

SQL
Python
R
AWS S3 / Athena
Snowflake
DuckDB

Product & AI

KPI design
Metric definitions
LLM evaluation
Roadmap input
Stakeholder demos

Visualization

Tableau
RShiny
Plotly
QuickSight
Looker
Streamlit

Sharyu Deshmukh

RWE Insight Copilot

NextRep

RapidAnalyzer

Publications

Education

Certifications

Statistics & causal inference

Data & engineering

Product & AI

Visualization

Building products where the evidence matters? Let's talk.