Apr 2022 – Aug 2025 · 3 yrs 4 mos

• Architected Lending Intelligence Platform from scratch — real-time AI covering acquisition, underwriting, disbursal, collections, retention, and cross-sell for a BNPL and cash-loan portfolio across multiple markets; defined AI product roadmap with PMs and ran A/B experiments measuring uplift in approval rate, fraud loss, and cure rate.
• Built BERT-based Decision Engine fusing SMS, banking, geo-location, app usage, bureau, transactions, and payment signals — enabling instant, explainable credit decisions and expanding access for thin-file and NTC customers.
• Designed and deployed modular LLM-Powered Fraud Framework: selfie/document/address/statement deduplication via embeddings and vector DBs (FAISS, Weaviate); XGBoost ensemble scoring across bureau, device, network, and behavioural signals; OCR-based forgery detection; red-teaming and DLP guardrails — achieving ~85% reduction in identity fraud.
• Launched Enterprise GenAI Assistant on Amazon Bedrock (Claude, Llama) with hybrid RAG (OpenSearch), multimodal document processing, DLP guardrails, and audit logging; designed GenAI enablement program reaching 100+ knowledge workers across Risk, Operations, and Marketing.
• Delivered CLTV model integrating spend, retention, and risk signals for precision targeting; built collections intelligence (intent-to-pay segmentation, next-best-action) improving cure rates and reducing cost-to-collect.
• Built and scaled AI organisation of 10–15: Data Scientists, ML Engineers, Product Managers. Defined roles, ran hiring, drove GenAI adoption through agent development, prompt engineering (CoT/ReAct), and business process redesign.
• Implemented enterprise AI Governance and MLOps: drift monitoring (PSI/CSI), SHAP/LIME explainability, MLflow model registry, CI/CD, champion-challenger testing, data contracts, lineage tracking, and LLM eval harness — regulatory-ready across GDPR and AML/KYC jurisdictions.

Dec 2020 – Mar 2022 · 1 yr 3 mos

Jun 2020 – Apr 2022 · 1 yr 10 mos

• In association with Great Learning

May 2020 – Nov 2020 · 6 mos

• Built Geo-Spatial Intelligence Platform using OpenStreetMap, DBSCAN density clustering, and NLP address normalisation — detecting location-based fraud rings, predicting address changes, and improving field collections routing.
• Designed address data pipeline with quality scoring, regional normalisation, and completeness validation — reducing operational errors and improving customer contactability across a distributed lending footprint.

Apr 2020 – Apr 2021 · 1 yr

Oct 2018 – May 2020 · 1 yr 7 mos

• Built large-scale Data Feature Lake integrating customer, network, and device data for 400M+ subscribers — unified 360° feature views enabling churn reduction, IoT analytics, and network intelligence across Jio's telecom, retail, and petroleum businesses.
• Engineered distributed ML pipelines on India's largest Hadoop cluster (MapReduce / Spark): real-time anomaly detection, demand forecasting, and network tower placement optimisation using geo-location and IoT sensor data.

Jul 2018 – Present · 7 yrs 9 mos · Hyderabad, Telangana, India

Jan 2018 – Sep 2018 · 8 mos

• Developed NLP-driven resume classification models using probabilistic (Naive Bayes), deep learning (LSTM), and rule-based methods to map candidate profiles to job descriptions.
• Applied text mining techniques including tokenization, word embeddings (Word2Vec), POS tagging, NER, sentiment analysis, and topic modeling to improve job-candidate matching.
• Designed and implemented a custom annotation platform for large-scale corpus tagging and supervised learning pipelines.
• Automated recruitment workflows by integrating candidate data from ATS and CRM systems, improving talent alignment and recruiter efficiency.

Dec 2015 – Jan 2018 · 2 yrs 1 mo · India

• Delivered end-to-end ML solutions, including data ingestion, cleaning, feature engineering, model development, validation, and deployment for large enterprise datasets.
• Applied statistical and ML modeling techniques (regression, k-NN, neural networks) to support click fraud detection use cases in digital advertising.
• Founded and led an internal ML Club, training 100+ junior engineers in Data Science, Machine Learning, and IoT technologies.
• Developed and maintained data ingestion pipelines, ensuring clean, scalable datasets for downstream modeling and analytics.

Aug 2013 – Oct 2015 · 2 yrs 2 mos · Chandigarh, Chandigarh, India

• Researched face recognition using symmetry-based features and classification via SVM and tree-based models.
• Developed image forgery detection techniques using block-wise segmentation and noise pattern analysis for morphed image identification.
• Built medical image analysis models for breast cancer detection through feature extraction and dimensionality reduction.