Democratizing institutional-grade financial diagnostics, portfolio auditing, and SEBI regulatory intelligence with zero cloud subscription fees. Built on a deterministic quant engine, serverless vector search, and Google Gemini.
Automated multi-bank CSV statement preprocessing pipeline. Cleans debit/credit columns, standardizes date formats, eliminates corrupted rows, and handles varying bank export schemas.
Rule-based taxonomy mapping ambiguous bank narrations into 7 distinct financial buckets (Housing & Utilities, Groceries, Shopping, Travel, Entertainment, Healthcare, Investments).
Two-tailed Gaussian distribution model ($Z = \frac{x - \mu}{\sigma}$) isolating abnormal transaction spikes from baseline category averages to flag one-off spending shocks.
Interactive visual dashboards delivering monthly income vs outflow comparisons, category expense doughnuts, net savings accumulation, and personal savings rate percentages.
Structured tool-calling pipeline with a 1M+ token context window. Employs zero-hallucination guardrails, fast-fail 429 quota handling, and SEBI disclaimers modeled after Groww G.1.
Deterministic SQL rule engine with Row-Level Security (RLS). Houses official exit load schedules, statutory asset allocation mandates, and user audit session persistence.
Serverless embedded vector store reading Parquet indices over Cloudflare R2 with $0 egress fees. Fast approximate nearest neighbors (ANN) across 500+ mutual fund factsheets.
Zero-disk in-memory PDF parsing via Python io.BytesIO streams. Instantly digests password-protected CAMS/KFintech CAS statements and multi-format bank PDFs.
Newton-Raphson multi-cashflow XIRR solver with SEBI <180d short-vintage guards, pairwise stock overlap Venn matrices, rolling 1Y/3Y alpha, and Section 112A/50AA tax math.
Stateless Python WSGI serverless deployment on Vercel with automated GitHub Actions offline batch indexing pipelines and continuous pytest suites.
Designed and engineered the core Financial Spending Analyzer framework. Built the end-to-end bank statement parsing pipelines, category classification engine, expense trend heuristics, and statistical spending anomaly detection algorithms.
Architected and implemented the Mutual Fund Intelligence Layer, Tri-Hybrid RAG routing pipeline, Newton-Raphson XIRR quant engine, SEBI short-vintage compounding baseline, Budget 2024 taxation engine, and the interactive Claude/Groww G.1 visual Chatbot advisor.
FinWise is strictly a software engineering portfolio, educational demonstration, and open-source quantitative research project. None of the features, diagnostic outputs, AI conversational responses, financial health scores, or automated budget categorizations constitute financial, investment, legal, or tax advice.