MNNIT Allahabad 2025 · Currently at Shriram Finance · Building AI systems for real business workflows.
I work in portfolio analytics at a large NBFC. Most of what I've built here started as a real problem I ran into at work, information that was hard to access, workflows that were manual, decisions that needed better data and are real operational bottlenecks. I build to solve those problems, not to demonstrate technology.
Open to Applied AI, AI Engineering, and Data Analyst roles.
🛠️ Tech Stack
- Upload portfolio data and ask questions in plain English
- Generate portfolio summaries instantly
- Surface collection insights and trends
- Reduce dependency on manual reporting workflows
Collection teams often spend significant time gathering information, analysing loan accounts, extracting information before making decisions. CollectionIQ reduces that friction by bringing insights directly to decision makers.
Tech Stack:
Python LangGraph Gemini Pandas Streamlit LangSmith
- 15 deterministic underwriting and risk validation checks
- Gemini-powered qualitative risk analysis
- Rule-based + LLM hybrid architecture
- RAG-powered borrower case Q&A
- Focused on explainable AI-assisted credit workflows
Python • Gemini • LangChain • PyMuPDF • Streamlit
- Reduced manual processing from 30–60 minutes to under 1 minute
- Supports CRIF High Mark and TransUnion formats
- Validation against report totals
- Gemini fallback pipeline for extraction correction
Python • PyMuPDF • Gemini API • OpenPyXL • Streamlit • LangSmith
Agentic content generation workflow built using LangGraph orchestration patterns.
- Planning + orchestration architecture
- Parallel worker execution
- Tavily-powered research pipeline
- Structured state sharing
- Evidence-driven content generation
Python • LangGraph • LangChain • Gemini API • Tavily
Production-style ML pipeline with CI/CD and AWS deployment.
- Modular ML pipeline
- Dockerized deployment
- GitHub Actions CI/CD
- AWS EC2 + ECR + S3 integration
- FastAPI serving layer
FastAPI • Docker • AWS • MongoDB • GitHub Actions
- Applied AI Systems
- Agentic Workflows
- Document Intelligence
- Retrieval-Augmented Generation (RAG)
- AI Reliability & Guardrails
- Underwriting Intelligence
- Collection Analytics
- Workflow Automation
- AI Product Systems