Rahul Sinha

Role Summary

Member of the Data Science Global Centre of Excellence team. I lead Data Science & Machine Learning initiatives in the Supply Chain area. I apply my Supply Chain domain knowledge & Data Science skills to execute global scale machine learning projects in functions like Procurement, Supply Chain Planning & SHE (Safety, Health & Environment). My current mandate is to explore & innovate algorithms applicable to Unilever's use cases through PoCs and lead them through to industrialization, build in-house expertise in niche but high impact areas of Data Science (for e.g., Graph Analytics & Reinforcement Learning) & facilitate Hiring, Product & Vendor assessment, Knowledge Imparting & Mentoring activities.

Sample Projects

• Automated Price Forecasting for Tail Materials (Procurement): Designed, built & deployed a Deep Learning based solution (along with rest of the workflow) to forecast prices of 200,000+ Tail Raw & Packaging Materials which saved 20,000 hours of effort annually for 750+ buyer community. Accuracy of forecasts went up from 80% - 85% range to 90% - 95%.
• Demand Forecast Engine (Demand Planning): This is a long running initiative at Unilever which aims to deliver the best possible volume forecasts for hundreds of thousands of SKU-Customer combinations across markets on a weekly basis. I have built & deployed a suite of models using Traditional ML (mostly Gradient Boosted Trees), Deep Learning (using customized Transformer Architectures & others) & Graph Neural Networks for several European & North American markets. Over the years, these models have delivered 20% or more accurate forecasts than the enterprise planning systems & Demand Planners.
• SHE Insights - Incident Prevention (Safety, Health & Environment): The project is aimed at safety managers to help them spot & mitigate high risk factors for injuries in Factories, Warehouses & Distribution Centers. I designed & built the solution using Bayesian Networks based on our Global Incidents database. The solution models various causal relationships among incident variables & identifies combinations of factors associated with a high probability of incident. Such insights help SHE teams to conduct required audits & institute SOPs to minimize safety risks.
• E-Commerce Demand Forecasting (Demand Planning, New Business Models): Started in the aftermath of Covid-19, this project caters to the unique characteristics of e-commerce demand with which the usual forecasting algorithms tend to struggle. This multi-stage probabilistic model built using proprietary Deep Learning architecture (a variant of the vanilla Transformer model) delivered an immense improvement in Bias ranging from 50% - 200%.
• Reinforcement Learning as alternative to Mathematical Optimization for Inventory Optimization (Supply Planning): Demonstrated remarkable RL capabilities in inventory planning by building a version of the 'Beer Game' allowing human players to compete against pre-trained RL agents. The RL models showed impressive performance outscoring the manual planners consistently & exceeding even black-box optimization methods. This is an active project under development.
• Graph ML for Forecasting (Planning, Graph CoE): Developed a handy library using Pytorch & PyG to showcase forecast improvement potential by explicitly incorporating inter-relations between various forecast keys. The library also provides comprehensive model explanation capabilities. This is an active project under development.

Role Summary

Ushered in Machine Learning based processes to replace extant ways-of-working & planning systems (SAP APO). BAU work included conducting proofs-of-concept driven through vendors and own contributions, & liaising with Data & Analytics organization to formalize & roll-out proven solutions in scalable manner.

Sample Initiatives

• Introduced for first time at Unilever, open-source Deep Learning architectures like DeepAR (by Amazon), TCN & MQRNN to circumvent the shortcomings of Traditional ML models vis-a-vis highly seasonal & volatile categories.
• Designed a solution to identify the impact of Quality Incidents on sales loss. This was done as part of an internal Unilever-wide competition & the solution won the Chief Quality Officer's award.

Role Summary

As IT Manager within the Enterprise & Technology Operations organization, now UniOps, ensured smooth functioning of Key Financial System SAP ECC & other integrated systems for Europe. In the latter half of engagement in UniOps, took on more functional roles as Planning & O2C SME.

Salient Projects

• Project Manager - SAP ECC Landscape (Europe) Upgrade & Migration: Delivered a highly invasive change affecting 21 markets through a rigorously planned & executed project over a period of 8 months & a budget of €2MM.
• Project Manager - Archiving: Led the implementation of OpenText Archiving system for global usage. Also implemented SOP for archiving key Business Objects for Europe resulting in 50% smaller data footprint, significant improvement in response-time & cost-savings through decommissioning of legacy systems.
• A key member of several regulatory projects such as Grexit, Brexit & sale of Unilever's Spreads business to KKR, the new entity now known as Upfield.

Role Summary

Member of Global Consulting Practices Team. Consulted in the capacity of Oracle Database expert with specialization in Administration, HAA & Performance Tuning.

Salient Projects

• Completed large scale Datacenter Migration & Consolidation project for Barclays Capital in wake of its acquisition of Lehman Brothers. Based out of London, UK, lead team of DBAs in India, UK & Singapore to move Lehman Brother' legacy systems on to Barclays compatible database infrastructure while upgrading Barclays own databases to latest supported versions with minimal downtime. Project was executed over 18 months from multiple sites, impacted over 1000 databases & required extensive stakeholder communications in addition to hard technical work.