Agentic AI
Orchestration, investigation workflows, and systems that reason with evidence.
LangGraph · Vertex AI · LLM-as-JudgeHi, I’m Prateek.
I build AI for
the real world.
A data science leader who still loves to build. I turn complex business problems into machine learning and generative AI products people use.
Senior Data Science ManagerLG Electronics North America
Experience & education
- Connected model capabilities
- 4
- Customer & product intelligence at LG
- KPIs in the monitoring catalog
- 26
- Forecasting, detection and AI investigation
- Student projects reviewed
- 1,000+
- Mentoring through Udacity
Selected work
Complex problems.
Useful products.
From detecting an unexpected business change to understanding a customer’s next purchase. A closer look at what I build.
Customer & product intelligenceWho to reach.
What to recommend.
I lead and build connected ML products at LG, turning customer and catalog signals into decisions people can use.
Four complementary capabilities. One story of applied data science, product thinking, and production delivery.
Next-basket recommendationsWhat will this customer want next?Read case study Customer lifetime value & purchase propensityWho is ready to buy, and where is the value?Read case study Category spend & price preferencesWhat price level fits this purchase?Read case study Explainable product alternativesWhat is a useful alternative, and why?Read case study
Real-time performance monitoring
Forecast & detect
Investigate
Explain & verify
Prioritise & act
From a business question to a working system.Workflow overview
LG Electronics
Finding the signal. Explaining the why.
A production platform that watches LG.com's commerce performance around the clock, investigates important changes automatically, and turns them into evidence-backed cards that business teams can act on.
- forecast-based detection
- Hourly
- typical investigation
- ~4 min
More from the workbench
Semantic search at AmazonAmazon NLP & search38,000+employee audienceHow I work
The model is
only the beginning.
I care about the full journey: the business question, the engineering, the evaluation, and what happens after launch.
Production ML
Forecasting, customer intelligence, and models connected to real business decisions.
Python · Time series · Gradient boostingGenerative AI
Grounded generation, contextual search, and human review built into the workflow.
RAG · Large language models · NLPThe journey so far
Always learning.
Always building.
From business analytics to leading production AI. Each role has brought me closer to the point where data becomes a decision.
View full résuméAug 2023 — PresentLG ElectronicsSenior Data Science Manager / Lead Data Scientist
New Jersey, United States
Lead the data science team, set technical direction, and personally design, code, and ship production AI systems. Own the work from discovery and modeling through deployment and adoption.
- Real-time commerce monitoring: Chronos-2 forecasting, hourly and daily anomaly detection, and AI investigation across a 26-KPI catalog.
- Production customer intelligence spanning next-purchase recommendations, purchase propensity, customer value, and category spend.
- Explainable product alternatives combining catalog context, business rules, and AI-assisted ranking.
- RAG-based campaign creation, LLM personalization, and an internal analytics assistant.
Jun 2022 — Mar 2023AmazonData Scientist
Seattle, United States
Built semantic search for an internal tool serving 38,000+ employees, and developed Python Dash applications to put ML and NLP capabilities in the hands of non-technical users.
- LSI, TF-IDF, SVD, hyperparameter tuning, and ensemble result ranking.
Jun 2019 — Oct 2020Fractal AnalyticsData Science Consultant
Mumbai, India
Worked with General Mills' finance team and Mondelez's global supply chain team on decision-support dashboards, business KPIs, and Python-based data automation.
- Built Tableau dashboards for finance, supply chain, and executive decision-making.
Mar 2019 — Jan 2021UdacityData Science Mentor & Reviewer
Remote
Mentored Business Analytics and Predictive Analytics Nanodegree students through live teaching and individual project reviews.
- Reviewed more than 1,000 student projects.
Oct 2018 — Feb 2019FabHotelsBusiness Intelligence Analyst
Gurugram, India
Worked with the founding team and CMO on booking reliability, fraud detection, and pricing. A customer scoring and confirmation process helped reduce no-show rates from 25% to 13%.
- Built regression models, Python analysis, and tools for cluster managers.
Nov 2017 — Jul 2018OneDirectBusiness Analyst
Bengaluru, India
Analyzed brand performance and marketing campaigns, built analytical reports, and introduced Python web crawlers to improve the team's data collection.
Jul 2017 — Aug 2017SpoylData Analyst Intern
Bengaluru, India
Supported acquisition and marketing teams with campaign analysis, data extraction, and dashboards.
A little about me
Curious by nature.
Hands-on by choice.
I’m happiest at the intersection of a hard problem, a blank editor, and a team that cares about getting it right.
At LG, I lead the data science team and stay close to the code. My work spans classical machine learning, multi-agent systems, and generative AI—with a focus on making them reliable enough for everyday use.
That means thinking beyond model accuracy. It means benchmarking, monitoring, and working with the people who will actually use what we build.
Boston University
Master’s in Applied Data Analytics
2021 – 2022Udacity
Machine Learning, Data Visualization &
Predictive Analytics Nanodegrees
Beyond the day job
Teaching is part
of the work.
I mentor through KaggleX and previously taught and reviewed more than 1,000 student projects at Udacity. I also write about data science, NLP, and machine learning.
Read my writing on MediumLet’s connect
Good things start
with a conversation.
Production ML, agentic workflows, or the gap between AI research and real products—I’m always happy to compare notes.