AI-Powered Email Automation
Intelligent microservice classifying financial emails with 95% accuracy. Extracts PAN, ARN codes using LLM zero-shot learning.
- Role: ML Engineer
- Timeline: July 2024 - September 2024
- Team: Solo + Integration Team
- Technologies: Python, Google Cloud, FastAPI, LLM, vLLM, Docker, REST API
- Link: https://github.com/Laxman824/
Problem
CAMS receives 1,000+ emails daily from mutual fund distributors, investors, and AMCs. Each email needs classification (complaint, query, transaction request) and extraction of key identifiers (PAN, Folio, ARN) for routing. Manual processing was slow and inconsistent.
Solution
Built an AI microservice using LLMs for zero-shot email classification and named entity extraction. The system classifies emails into 15+ categories, extracts structured data (PAN, dates, amounts), and routes to appropriate teams - all without training data.
Impact
- 95% classification accuracy
- Processing 1,000+ emails daily
- 80% reduction in manual work
- Sub-second response time
Key features
- Zero-shot email classification (15+ categories)
- PAN, ARN, Folio number extraction
- Date and amount detection
- Sentiment analysis for prioritization
- Confidence scoring for human review
- REST API for integration
- Batch processing support
- Audit logging for compliance
Tech stack
- Backend: Python, FastAPI, Pydantic, asyncio
- Ai: vLLM, Mistral-7B, LangChain, Instructor
- Infrastructure: Docker, Kubernetes, Redis
- Monitoring: Prometheus, Grafana, Sentry
What K Laxman learned
- Deploying LLMs in production with vLLM
- Structured output generation with Instructor
- Building high-throughput async APIs
- Zero-shot vs fine-tuned model tradeoffs
Explore more
- Home — overview, skills and a built-in AI assistant
- Experience — roles at Think360 AI (CAMS), CAMS Mutual Funds and IIT Delhi
- Projects — GenAI, LLM, RAG and full-stack builds
- Education — IIT Delhi, M.Tech & B.Tech Computer Science
- GitHub Activity — open-source contributions
- Contact / Hire me