arun@portfolio:~$ whoami_
I engineer retrieval systems, local LLM workflows and AI-assisted products—then build the APIs and interfaces that make them genuinely useful.
I'm interested in the hard parts of applied AI: how context is retrieved, how outputs are evaluated, how privacy is protected, and how a model becomes a dependable product.
01 / SELECTED AI WORK
Experiments and products focused on grounded outputs, transparent reasoning and useful human control.
LOCAL AI · RESPONSIBLE ML
A privacy-first screening workspace that compares resumes with a job description, then surfaces ranked evidence and gaps for human review. Local Ollama inference adds structured fit analysis alongside an explainable deterministic score.
// Resume text stays local by default. No cloud API or account required.
VIEW SOURCELLM · DEVELOPER TOOLING
A code intelligence assistant for C/C++, Java and Python that retrieves relevant context before explaining, debugging and optimizing code, with summaries, suggestions and unit tests.
// Designed around retrieval quality and code-grounded responses.
RAG · KNOWLEDGE SYSTEMS
A PDF question-answering system that turns documents into cited, context-aware conversations using chunking, vector embeddings, semantic retrieval and persistent chat history.
// Built to make answers traceable to source material.
APPLIED AI · MOBILE SAFETY
A cross-platform safety product combining AI-assisted threat assessment with SOS workflows, live location, evidence capture and native Android services.
// Exploring AI where latency, reliability and safety matter.
02 / ENGINEERING RANGE
Strong AI products still need well-shaped APIs, reliable persistence and interfaces that communicate clearly.
AI / MLLangChain, Hugging Face, Ollama, TensorFlow, PyTorch, scikit-learn
RETRIEVALEmbeddings, semantic search, FAISS, ChromaDB, chunking, citations
BACKENDPython, FastAPI, Node.js, Express, REST, SQL, MongoDB, SQLite
PRODUCTReact, JavaScript, Flutter, Dart, Kotlin, Android, Git
JWT authentication, CRUD workflows, REST APIs and responsive task handling.
React · Node.js · Express · MongoDB AtlasReusable Flutter UI integrated with GPS, maps, media capture, persistence and native services.
Flutter · Dart · Kotlin · AndroidLong-range secure push-to-talk voice communication using ESP32 and LoRa.
ESP32 · LoRa · Embedded systems03 / OPERATING PRINCIPLES
I'm a Computer Science student specializing in AI at VIT-AP, focused on the full lifecycle of intelligent systems: preparing data, retrieving the right context, constraining model behavior, evaluating outputs, and shipping the experience behind a clean product interface.
My projects are how I test those ideas—local-first screening with human oversight, source-grounded document answers, code-aware retrieval and safety-focused mobile intelligence. Contributing to the VIT-AP Machine Learning Club also sharpened how I explain technical ideas to people.
PROGRAMB.Tech CSE · AI SpecializationVIT-AP University · 2023–2027
CGPA8.01 / 10Current academic record
APPROACHBuild · Evaluate · IterateResponsible, product-minded AI
04 / START A CONVERSATION