🚀 Ai Engineer | Data Scientist | LLM Engineer | Deep Learning & Transformers Enthusiast
I specialize in LLM Agents, Deep Learning, and Transformers, working on AI-driven solutions for fintech, analytics, and automation. My expertise spans NLP, ML models, and scalable data solutions that power intelligent systems.
📍 CARS24 | Gurugram, India | Mar 2025 - Present
✅ Aadhaar Masking Automation
- Built a blurriness detection model using Swin Transformer to assess scanned document quality.
- Automated Aadhaar number and QR code masking using RapidOCR, PyTesseract, and Google Vision API.
- Enhanced KYC automation with real-time PII redaction, reducing manual effort.
- Delivered ₹10L/month cost savings through efficient document processing.
✅ FCU (Fraud Control Unit) Automation
- Designed a fully automated FCU pipeline combining image quality checks and OCR-based data extraction.
- Built parallel Selenium automation APIs to validate user identity via government portals.
- Integrated Google Gemini API for real-time CAPTCHA solving in the verification flow.
- Improved fraud detection accuracy, resulting in ₹85L/year cost savings.
✅ Insurance Propensity Model
- Developed and deployed a CatBoost classification model to predict insurance purchase likelihood.
- Achieved a lift of 2 across deciles, enabling targeted sales interventions.
- Drove ₹50L/month incremental revenue impact through improved targeting.
✅ Feature Mart by Agentic AI
- Built an Agentic AI–powered feature mart with 10,000+ features across Bureau, Buy, Sell, Marketing, Calling, and Service data.
- Automated Snowflake queries, data flattening, and LLM-driven feature creation, cutting analytics turnaround by 40%.
- Delivered validated, LLM-generated features improving model accuracy and campaign targeting effectiveness.
📍 Legal-Pythia | Glasgow, UK | Jan 2025 - Mar 2025
✅ Chatbot Development for FinTech
- Built an AI chatbot using LangChain & RAG for financial queries.
- Integrated Hugging Face & Ollama for enhanced contextual understanding.
- Developed APIs for seamless integration with financial systems.
✅ Automated Reconciliation & Analytics
- Automated transaction reconciliation using BigQuery, SQL, and Python.
- Designed Looker dashboards for real-time insights and anomaly detection.
📍 BharatPe | Gurugram, India | Jul 2024 - Jan 2025
✅ KYC Stack (Name Matching with PAN & Aadhaar)
- Built a BERT-based name-matching model, improving KYC accuracy by 15%.
- Integrated Siamese BERT & FuzzyWuzzy for better transliteration handling.
- Automated real-time KYC workflows, reducing verification time by 20%.
✅ Loan Propensity Model
- Developed an XGBoost-based loan propensity model for 400,000+ merchants.
- Achieved AUC of 80-85%, increasing loan conversions by 6%.
✅ Speaker Notification System
- Built an NLP-powered notification system using Mistral & GLiNER for entity recognition.
- Replaced regex-based classification, significantly improving alert accuracy.
