About Me
I am a Data Scientist & ML Engineer at QuaEra Insights, specializing in deep-learning recommendation systems for personalized marketing and decisioning. I own the full ML lifecycle—from data ingestion and feature engineering to model development, rigorous evaluation, experimentation, and production deployment.
At QuaEra, I improved Aignyte's AI decisioning platform for direct-mail and email campaigns, delivering a 200% response-rate lift and ~50% gains across five clients. Earlier at Vedanta, I optimized procurement with forecasting models that saved $18M+ and boosted operational efficiency.
I hold a Master's in Data Science from the University of Rochester (4.0 GPA) and a B.Tech in Chemical Engineering from IIT Kanpur.
Experience
- Improved Aignyte's AI decisioning engine; delivered +200% lift in response rate for a Fortune 500 client
- Led five proof-of-value projects, achieving ~50% average improvement in response rates
- Built a proprietary "creative genome" system using LLMs (LangChain/LangGraph), XGBoost
- Managed end-to-end ML lifecycle on Azure Databricks and AWS
University of Rochester - Lopatkin Lab · Rochester, NY,
USA- Co-developed PlasAnn, a plasmid annotation tool; released via PyPI
- Developed reproducible bioinformatics pipelines using Python & R on Linux HPC cluster
- Co-authored peer-reviewed papers on plasmid prevalence and annotation methods
- Built a Python-based data analysis copilot using ChatGPT + fine-tuned LLaMA 3.3
- Designed a modular, agentic LLM framework and benchmarked 4 pipeline variants
- Productionized the top-performing stack; projected to save 200+ person-hours/month
Teaching Assistant for GBA465 - Python Analytics for R Programmers.
Vedanta Limited · Jharsuguda, Odisha, India- Optimized coal procurement using time-series forecasting; saved $18M+; received CEO Appreciation Award
- Developed power price forecasting model; boosted boiler efficiency 5%
- Improved aluminium product quality and domestic market share 3%
IIT Kanpur · Kanpur, UP, India
Created a custom vision algorithm to track colloidal particles; achieved ≈95% tracking accuracy.
KPIT Technologies · Pune, MH, India (Remote)Developed ML model to predict remaining engine oil life; XGBoost performed best (R² = 0.9).
Tomsk Polytechnic University · Tomsk, RussiaBuilt deep-learning application (ENet) for real-time semantic segmentation; achieved 67% mean IoU.
Publications
Skills & Tools
Focus Areas: Recommendation systems, personalization & decisioning, deep learning, LLMs & generative AI, experimental design & A/B testing.
- Python
- SQL
- PyTorch
- TensorFlow
- XGBoost
- LangChain
- MLflow
- Azure
- AWS
- Databricks
- Power BI
- Streamlit
- Git
- Docker