About
Arunesh Waran
AI, ML & Web Application Developer
B.Tech Artificial Intelligence & Machine Learning student building AI-powered applications end to end: retrieval pipelines, model integration, async APIs, database design, and containerized deployment.
Built a full-stack RAG assistant and a reliability auditor for AI agent workflows, both covered by automated tests in CI.
Eager to contribute to live AI, data, and automation projects.
Coimbatore, Tamil Nadu, India
Projects
AI Developer Assistant - RAG + LLM Full-Stack Platform
A full-stack assistant answering programming questions from official documentation and indexed GitHub repositories. Answers cite their exact source to reduce hallucinations.
LangChain, sentence-transformer embeddings, and FAISS/ChromaDB storage power hybrid dense + BM25 retrieval. A cross-encoder reranker supports semantic code search and stack-trace debugging.
Async FastAPI backend with PostgreSQL, Redis, API-key authentication, rate limiting, and health probes. Next.js + TypeScript UI, Docker Compose deployment, and CI gates.
Tools: Python, FastAPI, LangChain, FAISS/ChromaDB, PostgreSQL, Redis, Next.js, TypeScript, Docker, GitHub Actions
View repository (opens in a new tab)
ReplayGuard - Reliability Auditor for AI Agent Workflows
A zero-dependency Python CLI that crash-tests AI agent workflows performing real side effects. Records intent before execution and reconciles against external truth to report duplicate or lost effects.
SQLite WAL and synchronous=FULL underpin idempotency and recovery, with cross-process leases, dead-holder reclamation, three replay policies, and process-death fault injection.
Property-based and concurrency tests across Python 3.11-3.13 on Linux, macOS, and Windows in GitHub Actions. Maintained 90%+ coverage with strict mypy and ruff gates.
Tools: Python, SQLite, CLI tooling, pytest, Hypothesis, mypy, GitHub Actions
View repository (opens in a new tab)
Smart Traffic Monitoring System - Computer Vision & Automation
Python computer vision that automatically detects and tracks vehicles in aerial footage to analyze traffic flow.
Built and tested deep-learning object detection in reusable detection, tracking, and reporting modules. Automated the workflow to surface vehicle-flow insights for real-time decisions.
Detection workflows use RF-DETR and YOLOv8.
Tools: Python, OpenCV, TensorFlow, YOLOv8, RF-DETR, Deep Learning, Data Analysis
View repository (opens in a new tab)
Skills
Languages
Python, TypeScript/JavaScript, SQL
AI / ML
RAG, LLM Integration, LangChain, Prompt Engineering, NLP, Machine Learning, Deep Learning, Scikit-learn, PyTorch, TensorFlow, OpenCV
Web Development
FastAPI, Flask, async Python, REST API design, Pydantic, Next.js, React, Tailwind CSS
Databases
PostgreSQL, MySQL, MongoDB, SQLite, Redis, SQLAlchemy, FAISS, ChromaDB (vector search)
Testing & DevOps
pytest, Vitest, Hypothesis, mypy, ruff, Docker, GitHub Actions (CI/CD), Git, Linux, Postman
Core Competencies
Python programming, LLM & RAG systems, machine learning & NLP, backend & REST APIs, full-stack web development, SQL & NoSQL databases, testing & CI/CD, data structures & algorithms.
Credits & Sources
An independent portfolio, not affiliated with or endorsed by Nokia. Nokia is a trademark of its respective owner.
Handset Photograph
Nokia 3310 Blue R7309170 (retouch) by Rainer Knäpper (smial), retouched by Multicherry, used under the Free Art License 1.3. Resized and converted to WebP, with a replacement LCD and interactive control overlays. The adapted handset visual is available under the same license. Download the adapted image.
Pixel glyphs, synthesized key tones, and the Snake II and Ask Arun apps are original work. Lucide icons use the ISC license. No original Nokia ringtone or proprietary font is redistributed.
Hardware References
Photographic perspective, regional key legends, and a slightly expanded LCD viewport are retained or adapted for usability. The rendered body is not a calibrated engineering model.