Avnish Vishwakarma

• Enhanced machine learning model performance by implementing advanced algorithms and feature engineering techniques.
• Collaborated with cross-functional teams to develop cutting-edge machine learning applications for various industries.

Leadership and Collaboration

- Assumed the role of a technical mentor, fostering high levels of productivity

among team members.

- Worked very closely with the CTO, customer success team and customer-

side POCs

- Collaborated across teams for product deliverables

Technology

- Led research and improvement for primarily three areas: Time series

forecasting algorithm, Post-Processing of raw model forecasts and large-scale

data processing.

Page 1 of 4- Led time series model tuning process for several clients.

- Developed and implemented techniques to scale the time series algorithm for

memory and time efficiency.

- Optimized data ingestion, processing and export pipelines to handle large

scale data.

- Guided in setting up inventory planning for clients

Time Series Forecasting

- Enhanced the time series forecasting algorithm by incorporating tree-based

forecasting methods.

- Implemented support for external signals, such as price, out of stock flag,

holiday, and weather data, to

improve the accuracy of the forecasts.

Forecast Post Processing

- Built an algorithm for post processing of raw forecasts to improve forecast

stability and accuracy.

- Added support for forecasting new products.

- Built a module to quantity the impact of holidays, events and promotions on

forecasts using historical

demand data.

Successfully deployed this algorithm for over 25+ customers spanning diverse

domains, including retail, e-commerce, D2C/CPG, restaurant chains, beauty &

cosmetics, and pharmaceuticals.

- Evaluated different timeseries algorithms including ARIMA, Winter Holts,

BSTS, STL and Croston.

- Developed an automated tuning process for these algorithms that resulted in

significant improvements in forecast accuracy.

- Built the core algorithm for timeseries forecasting for SKU - Warehouse-Store combination at daily, weekly and monthly granularity.

- Implemented ensembling techniques to further improve forecast accuracy.

1. Created an end to end pipeline for automating the ELT (Extract, Load and

Transform) operation

of data from flat files into the Hadoop ecosystem using Shell Scripting and SQL

(Cloudera

Impala).

2. Created reports and dashboards in Power BI for C-Suite level on Desktop as

well as mobile

platform on top of views and tables in Hadoop ecosystem.

3. Created reports and dashboards in Power BI for Finance team with custom

functionality on top

of views and tables in Hadoop ecosystem.