Aug 2024 – Dec 2024 · 4 mos

• Taking care of my first born

May 2023 – Aug 2024 · 1 yr 3 mos · Bengaluru, Karnataka, India · Hybrid

• Successfully managed and mentored a team of 5 senior and junior data scientists promoting a collaborative and high-performing work environment
• Seeker Search:
• Migrated seeker search from classical BM25 based lexical matching to Hybrid Search using all-mpnet-base-v2 model to overcome limitation of “Term Frequency” in BM25 and thereby increase relevance in search results
• Hybrid search led to increase in average no. of results per query by 52% and reduction in sparse results by 9%, significantly enhancing user experience
• Seeker Registration Flow - Enhancement:
• Improved the consumption of resume parser by replacing fuzzy matching on master dataset with SBERT all-MiniLM-L12-v2 Bi-encoder based similarity algorithm leading to increase in fill rate by 32% at an overall level, streamlining the registration flow and improving data accuracy
• Profile Matching:
• Developed two step methodology to identify duplicate profiles in two databases by leveraging Elastic Search BM25 algo to reduce search space to ~0.01% keeping recall at ~51% and matching profile attributes with high precision of ~93% using USE multilingual large V3 encoder
• Model led to reduction of Unified database by 4.34% by matching same candidate profiles in individual databases
• Prep AI:
• Integrated OpenAI Text-DaVinci model (GPT 3.5) to generate personalised interview questions for each seeker-job pair at scale serving 200k apply/day leading to increase in repeat customer engagement by 3.4%

Dec 2021 – Jun 2023 · 1 yr 6 mos · Remote

Jul 2021 – May 2023 · 1 yr 10 mos · Bengaluru, Karnataka, India · Remote

• Merchant Highlights:
• Deployed an information retrieval pipeline for dish names and key phrases using dependency parser based rules along with their respective sentiment with ~ 94% accuracy using Word2Vec embedding based XGBoost Classifier.
• The treatment group with new UX based on the above mentioned items showed an improvement of 0.4% in CTB (Click to book ratio) and 3.4% in CTR (Click through rate)
• Recommendation Engine - Candidate Generator :
• Developed a two tower neural network architecture retriever based on user and merchant embeddings to narrow down the candidate pool the re-ranker model by 86% and with recall of 71%

Sep 2020 – Jun 2021 · 9 mos

• Anomaly Detection:
• Created a generic framework to detect anomalies in real-time using adtk library by Arundo.
• The framework was built with the intention so that all internal pods can start alerting mechanisms for their respective metrices and react quickly.
• Cancellation Prediction :
• Predicted the absolute number of orders getting cancelled by customers and identified major drivers leading to it using Random Forest Regressor algorithm.
• The model was developed for the purpose of overbooking and has shown an increase of 2.7% in capacity utilisation during the pilot phase

Aug 2020 – Sep 2020 · 1 mo

Apr 2019 – Jul 2020 · 1 yr 3 mos

• Actual Capacity Estimation :
• Quantified ‘effort factor’ for delivering an actual order by considering ideal order as the benchmark using CatBoost Regressor algorithm considering geographical distance and order data like items per order, weight, item types etc. as dependent variables.
• Productionising the mentioned model led to reduction in reschedule by 21% along with approximate decrease of Rs 1.7 in cost per order
• Promise Time Simulation:
• Simulated promise time with and without incentivising delivery time slots based on FIFO method using Python to support future capacity planning
• Built a framework for predicting the optimal time slots to incentivise and cashback values respectively to reduce the promise time for the customers wanting earliest possible delivery

May 2018 – Mar 2019 · 10 mos

• Last mile operations optimisation:
• Interacted with field operatives to identify their major pain points and initiated better data capturing
• Created an interactive dashboard using 'Apache Superset' to provide data visibility of complaints, reschedules and cancellations at the most granular level
• Reduced the complaints by ~ 4% by conducting multiple experiments and monitoring their impact by using A/B testing and pre-post analysis

Nov 2016 – May 2018 · 1 yr 6 mos · Banglore

• Incentive Compensation, Goal Attainment & Goal Setting:
• Helped a leading US Pharmacy Retailer set goal targets across different quarters, analyze goal attainment and set Incentives & Payouts for their sales force
• Created automated framework using SAS & Excel VBA to support goal setting & payout calculation that enabled faster decision making by bringing down the time taken to test out various scenarios by 75%
• Sales Forecasting:
• Forecasted monthly revenue for a fortune 50 Client to facilitate better quarterly goal setting that ensured accurate payouts during disaster hit months
• Best forecast model was chosen for different panels using ARIMA, S-ARIMA or Holt Winters model
• Prescriber targeting:
• Optimised costs of prescriber targeting for a fortune 500 pharmacy retailer by identifying high value targets and generating a projected positive EBIT of 60M $ over the next five years (starting FY16H2)
• Designed a flexible call plan tool for Sales Representatives bringing down Target Review time by 30%, CPT collation time by 50% and making call plans functional one week before FY start
• Market Potential Analysis
• Analyzed potential of various insurance plans to identify top contracts which will drive revenue and other top line performance metrics. Created an audience list of plan level exclusive targets which will be used for call planning
• Analyzed the potential of drugs across different regions and recommended stores where expansion would lead to high profitability

Jan 2016 – May 2016 · 4 mos · Gurgaon, India

Jun 2014 – Jul 2014 · 1 mo · Ghaziabad, Uttar Pradesh, India · On-site

• Project: An overview of power distribution system at 33 KV Substation