Mohit Dholi

About Mohit Dholi

Freedom. Responsibility. Curiosity.These have been my core assets. I have used these principles in every aspect of my life whether it be personal, professional, emotional, physical, friendship or family matters. Growing through challenging circumstances has shaped my life. The learning process has been phenomenal. I believe in my hard work, perseverance, nourishment of communication channels and never losing sight of my objectives.

Experience

1. Senior Data Scientist

   HARMAN International

   Sep 2020 — Present · Seattle, WA, US

   Exposing hidden relationships: Developed embedded graph models, linking 200k+ partners with 20+ nodalfeatures, to gain insights into the structure and hidden relationships within a region. The model is actively usedto determine a similarity score between a new entity and the old structure Converting sales pressure to risk: Developed ML models to detect sales target pressure in all the regions worldwide, quantify the target pressure to a collusive score, and detect corruption risk at a regional level. Validated the model against ground level third-party reports. Placed 3+ new regions under higher risk alert Sales team turnover: Developed ML models to measure risk associated with attrition in a sales team, within a region. The unbundling of attrition rates at multiple role levels, revenue and at different points in time provided meaningful insights into collusion practices. Among others, a synchronous attrition activity across close-by regions was a major discovery and revealed a new risk signal for the legal team to delve deeper into. Resource forecasting: Developed resource demand forecasting models for 5+workstreams for the vetting business team. Developed models using ARIMA, VAR, Random Forest, Autoencoders, FB Prophet and Orbit, and model selection was done based on least testing error principle. The error ranged between ±5% to ±17%. Concession abuse and imbalance: Developed ML models to discover patterns in which a partner may utilizethe 10+ concessions available to them. Derived abuse and imbalanced usage of concessions by partners withina region. Achieved 22%+ higher accuracy than the previous method in different regions. The model revealed$70M+ abuse of concessions Data extraction from external documents (Pilot): Finetuned Llama-3-8B-Instruct model on 8k+ external vetting reports (English) conducted on partners. A beta version of the front-end application on Gradio was released to the business team to extract data from the reports

Education

• Texas A&M University

  Master of Science (M.S.), Petroleum Engineering

  2012 — 2016

• Texas A&M University

  Doctor of Philosophy (Ph.D.), Petroleum Engineering

  2016 — 2020

• Indian Institute of Technology, Roorkee

  Bachelor of Technology (B.Tech.), Civil Engineering

  2004 — 2008

Skills

• Financial Modeling
• Rstudio
• Petrel
• Java
• Mathcad
• Visual Basic
• Fortran
• Ecrin
• Ms Project
• Rock Mechanics
• Staad
• Reservior Engineering
• Soccer
• Liasoning
• Badminton
• Structural Engineering
• Procurement Contracts
• Project Estimation
• R
• Autocad
• Labview
• Fekete Harmony
• Concrete
• Matlab
• Contract Management
• Drilling
• Arcgis
• Structural Analysis
• Civil Engineering
• Python
• Engineering
• Petroleum Engineering
• C++
• Production Engineering
• Mangrove
• Petrophysics
• Finite Element Analysis