Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

Backend-focused software engineer with hands-on experience building production systems in Java and Python, spanning SQL-heavy data modeling, transactional backends, batch processing, and retrieval-augmented systems. Proven ability to optimize system performance and cost (latency, token usage, query efficiency) and to design correctness-driven workflows for real-world domains including telecom billing audits, inventory intelligence, and ML-backed platforms. Experienced with RAG pipelines, vector search (FAISS/Chroma), cloud and on-prem deployments, and end-to-end ownership from backend architecture to APIs and debugging in production.…Backend-focused software engineer with hands-on experience building production systems in Java and Python, spanning SQL-heavy data modeling, transactional backends, batch processing, and retrieval-augmented systems. Proven ability to optimize system performance and cost (latency, token usage, query efficiency) and to design correctness-driven workflows for real-world domains including telecom billing audits, inventory intelligence, and ML-backed platforms. Experienced with RAG pipelines, vector search (FAISS/Chroma), cloud and on-prem deployments, and end-to-end ownership from backend architecture to APIs and debugging in production.

Sai Arvind Krishnan

Back-End Developer, Full Stack Developer, Web Developer, +7





Backend-focused software engineer with hands-on experience building production systems in Java and Python, spanning SQL-heavy data modeling, transactional backends, batch processing, and retrieval-augmented systems. Proven ability to optimize system performance and cost (latency, token usage, query efficiency) and to design correctness-driven workflows for real-world domains including telecom billing audits, inventory intelligence, and ML-backed platforms. Experienced with RAG pipelines, vector search (FAISS/Chroma), cloud and on-prem deployments, and end-to-end ownership from backend architecture to APIs and debugging in production.…Backend-focused software engineer with hands-on experience building production systems in Java and Python, spanning SQL-heavy data modeling, transactional backends, batch processing, and retrieval-augmented systems. Proven ability to optimize system performance and cost (latency, token usage, query efficiency) and to design correctness-driven workflows for real-world domains including telecom billing audits, inventory intelligence, and ML-backed platforms. Experienced with RAG pipelines, vector search (FAISS/Chroma), cloud and on-prem deployments, and end-to-end ownership from backend architecture to APIs and debugging in production.

Available to hire

Backend-focused software engineer with hands-on experience building production systems in Java and Python, spanning SQL-heavy data modeling, transactional
backends, batch processing, and retrieval-augmented systems. Proven ability to optimize system performance and cost (latency, token usage, query efficiency) and to
design correctness-driven workflows for real-world domains including telecom billing audits, inventory intelligence, and ML-backed platforms. Experienced with RAG
pipelines, vector search (FAISS/Chroma), cloud and on-prem deployments, and end-to-end ownership from backend architecture to APIs and debugging in production.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Language

English

Fluent

Hindi

Intermediate

Work Experience

Software Engineering Intern at SetuCS.io

August 1, 2025 - Present

Reduced Gemini RAG inference by consolidating multiple LLM calls into a single request, batching vector retrieval results, and parsing structured responses, reducing P95 latency by 35% and token usage by 40% in Node.js/Express using Axios and ChromaDB. Re-architected relational SQL schema to 3NF with FK constraints, composite indexes, indexed joins, and transaction-safe migrations, improving referential integrity and multi-tenant query performance by 25%. Deployed the platform on Azure VMs behind a NAT Gateway with Azure Monitor integration, improving mean time to detection by 40%. Implemented JWT authentication and Casbin-based RBAC to control API access by role and policy, reducing authorization-related defects by 30%. Delivered a performant React/TypeScript frontend (Vite, Tailwind, shadcn/ui, React Router, TanStack Query) and integrated RAG/ML endpoints, improving dashboard load times by 30%.

Software Engineer (Capstone) at TEOCO Corporation

January 1, 2026 - Present

Built Python-based PDF ingestion and extraction pipelines (pdfplumber / PyMuPDF), structuring telecom billing invoices for enterprise clients (AT&T, Verizon) into normalized PostgreSQL schemas for audit and analysis. Implemented Python validation/audit workflows with cross-document field matching, numeric consistency checks, and overcharge detection, emitting traceable outputs. Added an optional RAG layer by chunking text, generating embeddings, and indexing with FAISS to enable semantic retrieval, evidence lookup, and human-in-the-loop reprocessing via Python APIs.

Software Engineering Intern at Inox Tools Ltd.

May 1, 2023 - December 1, 2023

Replaced spreadsheet-based vendor and inventory tracking by building an on-prem Java 17/Spring Boot backend, modeling vendors, SKUs, purchase orders, goods receipts, returns, and inventory ledgers using PostgreSQL + Spring Data JPA. Implemented internal REST APIs (Spring Web) to expose vendor KPIs, inventory state, and procurement data directly from relational tables. Automated recurring operational workflows with Spring Scheduler to compute vendor performance metrics, inventory summaries, and exception reports, reducing manual reconciliation effort by 40% weekly. Introduced ledger-based inventory transactions, idempotent updates, and transactional boundaries to prevent double-counting. Improved backend reliability with Bean Validation, centralized exception handling, structured logging (SLF4J/Logback), and audit tables.

Research Intern at Vellore Institute of Technology (VIT)

June 1, 2021 - November 1, 2021

Built an end-to-end Music Information Retrieval (MIR) and remix pipeline in Python using librosa, NumPy, scikit-learn, SoundFile, argparse, and setuptools for modular ML+CLI workflows. Implemented MFCC, Chroma STFT, beat-synchronous segmentation, and statistical pooling to create fixed-length embeddings for similarity search. Developed pairwise similarity scoring (cosine, Euclidean) and embedding-space retrieval via kNN. Designed and trained a deep-learning RNN similarity model (sequence input, 0-1 similarity output) using TensorFlow/Keras or PyTorch. Productionized the prototype with PyTest, artifact pipelines, and reusable CLI commands for processing, feature extraction, model fitting, and remix generation.