Jan 2019 – Feb 2020 · 1 yr 1 mo · Gurugram, Haryana, India

• Projects :-
• Branch-Warehouse Intracity Route Scheduler: Identified vehicle composition and schedule for intracity material movement minimising cost and delays. Represented problem as time space network and formulated a Linear MIP
• Auto-planning consignments in trucks: Decided which consignments to send in truck for achieving best utilisation without increasing delay by formulating a genetic algorithm having multi objective functions; each time picking best from Pareto optimal front. Increased the weight utilisation of trucks by 150 basis points
• Flagging consignment for dimension entry mistakes: Created a Xg-boost model to do binary prediction if a consignment needs to be re-checked for dimension entry mistake. Dropped flag rate from earlier 60% to 16%. The solution had a direct impact on reducing DSO by about 10 days as post billing issues reduced

Mar 2018 – Dec 2018 · 9 mos · Gurugram, Haryana, India

Mar 2017 – Feb 2018 · 11 mos · Bengaluru, Karnataka, India

• Two Roads wanted to experiment new things in HFT through Data Science. Their vision was to run Neural network code for trading strategies. We started with automating the manual tasks that a trader did on day to day basis. Two of these tasks were picking which strategy to run daily and at what stop-loss. We researched them and were able to convert both these into an automated algorithms.
• Projects :-
• Trading Strategies using Machine Learning: Created a model to pick the best strategies to trade from the pool by predicting their PnL(profit & loss) using matrix factorisation and time series analysis
• Automatic Risk Allocation Module: Predicted how much risk (stop-loss) should each product be given for trading today. Allocated Risk by optimising combined sharpe of PnL(profits & loss) per unit risk subject to constraints. Backtested the module for last 2 years and showed it to be 100% more profitable than current manual allocation

Jul 2016 – Mar 2017 · 8 mos · Noida, Uttar Pradesh, India

• Opera solution was an analytics consulting firm. We used to create models and solutions for clients. Process/Project involved: - 1. Brain storming to arrive at optimal solution, 2. Interact with different teams (Business, Software and Analytics) from 3 different geographies (US, Shanghai and India), 3. Effectively execute to create a client deliverable solution, 4. Presenting results to client
• Projects :-
• Assess Risk of Project undertaken for a Legal Consulting Firm:
• For each prospective project, predict how risky it can be given which Partner from the firm is working, what services are being offered and to which Client. Along with a measure of risk, also pointed out where exactly risk exists so that the firm can make changes accordingly. It was an unsupervised problem . We developed a tree based structure to aggregate risk. Individual leaf risk was formulated using sigmoid functions, chi-squared tests and continuous feedback from clients.

Jul 2014 – Jun 2016 · 1 yr 11 mos · Noida, Uttar Pradesh, India

• Campaign model for credit card client: Predicted which customers are most likely to keep using promo card offered and their average spend on card using logistic and linear regression. Helped the client to run a targeted campaign