Experience.

Where I’ve worked.

1. Aug 2025 – PresentBengaluru, IN

   Gen AI Architect

   SBER Bank India

   • Fraud detection: led a team to architect a real-time fraud platform for money-mule detection, combining XGBoost, graph analytics and a GigaChat SLM investigation agent with SHAP explainability. It reached 93.5% PR-AUC at 4.25 ms latency and cut fraud losses by 76% compared with the baseline.
   • RAG knowledge platform: led a team to build a RAG-based AI assistant for 500+ employees across HR, IT and operations. It handles about 65% of queries without a human, cut response time from hours to seconds and saved 3 FTE.
   • Voice agent: led a team to build a production STT → LLM → TTS banking voice agent on Sber Platform V, with fine-tuned GigaChat 2 MAX, guardrails and a CI quality gate. It cut ASR WER from 46% to 22%, reached 92% intent accuracy and ran at 1.8 s p95 latency and ₹0.48 per turn.

   Real-time fraud detectionRAG knowledge assistantBanking voice agent

2. Apr 2024 – Jul 2025Bengaluru, IN

   Lead AI Engineer

   Ernst & Young LLP

   • Built an agentic AI analytics platform with LLM tool-calling, validated Text-to-SQL (about 88% execution accuracy) and auto-generated charts, so 300+ users could analyze sales, inventory and logistics data in plain language. It cut ad-hoc report requests by about 50% and turnaround from days to minutes.

   Agentic analytics platform

3. Apr 2022 – Mar 2024Bengaluru, IN

   Senior Data Scientist

   Société Générale

   • Built an LLM-driven onboarding platform that automates OCR-based document extraction, KYC checks and credit-risk scoring (0.87 AUC, with SHAP explanations), cutting turnaround from 3–5 days to under 24 hours and manual review by about 60%, with PII masking and on-premise Jenkins CI/CD.
   • Developed a production predictive model for justifying Net Asset Value (NAV) in investment management using decision trees, RabbitMQ, Python, AWS S3, PostgreSQL, Airflow, Docker, and Kubernetes.

   LLM-driven KYC onboardingNAV justification model

4. Feb 2019 – Mar 2022Bengaluru, IN

   Applied Machine Learning Engineer

   TCS Research

   • Built an NLP-based query chatbot (Django) for issue searching and a next-word-prediction module using an LSTM sequential model (Keras/TensorFlow/PyTorch).
   • Handled large-scale ML training and prediction on Hadoop/Spark distributed systems, with A/B testing and MLOps for lifecycle management.

   NLP query chatbot

5. May 2018 – Jan 2019Trivandrum, IN

   Applied Machine Learning Engineer

   GateOverflow

   • Performed semantic analysis on user comments using NLTK to extract customer sentiment and intent.