Open to opportunities

KOTAGIRI SATHWIK

@kotagirisathwik

Message

I build reliable backend and AI systems for financial workflows, document intelligence, and automation.

India

Message

What I'm looking for

I'm looking for opportunities to build backend and AI systems, especially reliable APIs, RAG applications, financial workflows, and production-ready automation.

I've built LedgerLens, an AI finance controller with a four-source reconciliation engine that processed 64,952 records per second and achieved 100% precision and recall on a 60-record benchmark.

For VERITAS, I built a high-performance PDF RAG platform using FastAPI, FAISS, LangChain, and LLaMA 3.3, with grounding controls, multi-tenant isolation, and fallback architecture for API rate limits.

I've also delivered FinSight, a real-time GST receipt analyzer that reduced invoice processing from five minutes to under three seconds. My work spans Python microservices, PostgreSQL concurrency, AI workflows, automated testing, and CI/CD deployments.

Experience

Work history, roles, and key accomplishments

LedgerLens — AI Finance Controller

Personal Project

Sep 2026 - Present (0 months)

Built a 4-source reconciliation engine matching payments, provider payouts, refunds, and bank credits with 100% precision and recall on a 60-record benchmark, processing 64,952 records/sec. Implemented deterministic Decimal arithmetic and SQLAlchemy optimistic concurrency control with audit trails.

fastAPI SQLAlchemy PostgreSQL React TypeScript Docker GitHub Actions CI CD Decimal Concurrency

VERITAS — High-Performance PDF RAG Platform

Personal Project

Aug 2026 - Present (1 month)

Built a document RAG platform using LangChain and LLaMA 3.3 70B with a cosine-similarity grounding floor to reject out-of-scope queries and eliminate hallucinated LLM generation. Reduced vector index cold-start latency to 2.76 ms by serializing FAISS embeddings to disk.

Python fastAPI Faiss Lang Chain LLaMA 2 Cosine Similarity Vector Databases Exponential backoff

FinSight — Real-Time GST Receipt Analyzer

Personal Project

Jul 2026 - Present (2 months)

Built an end-to-end financial extraction pipeline combining Vision OCR, Pydantic validation, and GST tax-rule verification for Input Tax Credit checks, reducing invoice processing from 5 minutes to under 3 seconds. Enforced multi-tenant data security using Supabase Row-Level Security policies.

Next.js fastAPI Groq Vision OCR Supabase Vercel Pydantic Playwright E2E Testing