Human-AI Collaboration Transformation
Transformed the engineering operational model for the company to an AI-first operating model reducing delivery time to half.
Sr. Technical Program Manager
Tell Me Your Story (TMYS)
Created a harness around video GenAI models for enforcing character, voice and scene continuity for long form videos
Masters of Business Administration
Customer Retention & Revenue Protection
Enabled early churn detection to protect high-value customer revenue by improving the performance of the base model from 77% to 97% using Gradient Boosting with class imbalance handling (under/over-sampling) and optimized hyperparameter tuning validated via cross-validation.
Predictive Pricing Engine
90% reduction in manual effort for pricing and eliminated the need to setup a new pricing team in a cash strapped start up.
Market Sentiment & Investment Analysis
Improved investment decision-making by uncovering weak sentiment–stock price correlation and preventing reliance on misleading signals, by building and evaluating multiple NLP models (Word2Vec, GloVe, Sentence Transformers), selecting a tuned GloVe model (F1 ~53%), and identifying key limitations such as overfitting, class imbalance, and prediction bias.
Senior Manager, Management Consulting
Imagination Labs
Kit Cat
A transcendental cat who will lead all of us to salvation. All bow to thee!
On-Demand Reporting & Analytics (NL2SQL)
Drastically reduced the time for custom requests, from weeks to minutes, and removed more than half of the engineering effort needed for these reports.
PG Diploma AI/ML
Conversational Knowledge Platform
Saved around half a million yearly in customer support and training costs.
Customer Targeting & Revenue Optimization
Established that advanced and costly pruning techniques don't provide better models in case of class imbalance in input data, and recommended the client to optimize the cost of implementing their product.
Imagination Labs
Dolma Doggie
A dog who's sole purpose is to tell us cheesy dad jokes. Beware!
Sr. Manager, Product Management
Plant Image Classification for Automated Weed Detection
Automated and improved seedling classification recall by 18 percentage points and performance by four percentage points by deploying an optimized CNN model with data augmentation and normalization techniques, enabling scalable and more accurate weed identification in agricultural workflows.
B. Engineering (Computer Science)
AI Product Leader with over 15 years across Amazon, Deloitte, and Startups. Building 0 → 1 products and 1 → scale using Agentic AI, MCPs, GenAI, LLMs, RAG, and ML to automate workflows, optimize decisions, and unlock value.