Skip to content
View Dev-2604's full-sized avatar

Block or report Dev-2604

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Dev-2604/README.md

Profile readme v2 · MD

Hi, I'm Deva Nanda Gopi 👋

Typing animation

B.Tech in AI & Machine Learning, Vidya Academy of Science and Technology, Thrissur (2026). I build applied AI systems — the kind where the model is one component of something that has to work end to end: an API, a database, a frontend, and a person on the other side of it.

Most of my work sits where LLMs meet real infrastructure — retrieval pipelines, FastAPI backends, and PostgreSQL, with a frontend that makes it usable.

Projects

ISL Interpreter — real-time Indian Sign Language to text and speech

MobileViT LSTM TensorFlow MediaPipe Flutter Flask

A Flutter app captures a gesture, a Flask server extracts MediaPipe hand landmarks, and two models classify it: MobileViT for the 35 static signs (98.8% accuracy) and an LSTM over 30-frame sequences for dynamic words (95.2%). Classifying landmarks rather than raw pixels keeps inference light and background-independent. Recognized signs are spoken aloud through TTS, with sub-second latency end to end.

InsightAI — AI-powered analytics platform

FastAPI PostgreSQL LangChain Mistral FAISS React Docker

Upload a CSV or Excel file and get an executive summary, KPIs, auto-suggested charts, and a chatbot that answers questions about your data. An LLM detects the header row in messy spreadsheets, a FAISS vector store powers retrieval-based Q&A, and JWT auth with per-user storage keeps uploads isolated. Deployed on Railway with PostgreSQL.

SevanaGPT — government scheme discovery for Indian citizens

RAG pgvector IndicTrans2 Mistral FastAPI Next.js Docker

India runs thousands of welfare schemes that most eligible citizens never claim, because the information is scattered and rarely in their own language. SevanaGPT puts 4,600+ schemes behind one conversational interface in 22 Indian languages — hybrid keyword + vector search fused with Reciprocal Rank Fusion, an FSM-driven chatbot to keep LLM behaviour predictable, and a self-hosted IndicTrans2 translation service.

Experience

  • IoT & Robotics — Stem Robotics, Kochi (June 2025) Obstacle-avoiding robot in embedded C++ with Arduino and ultrasonic sensors.
  • AI & Python — Pacelab, Kochi (June–July 2024) Built a Python chatbot; designed conversation flow and response-accuracy tuning.
  • Cybersecurity (virtual) — Cisco Networking Academy (May–July 2024) Network security, vulnerabilities, and risk management fundamentals.

Toolkit

Languages Python · C · R · Dart · SQL AI / ML TensorFlow · LangChain · Mistral · RAG · FAISS · pgvector · MediaPipe · OpenCV Backend FastAPI · Flask · PostgreSQL · SQLAlchemy Frontend React · Next.js · Tailwind CSS · Flutter Tooling Docker · Git · Linux

Reach me

devanandakgopi@gmail.com · LinkedIn · Sharjah,UAE

Pinned Loading

  1. InsightAI InsightAI Public

    AI-powered analytics dashboard: turn CSV/Excel uploads into executive summaries, KPIs, charts, and a chat-with-your-data assistant.

    JavaScript

  2. ISL-Interpreter ISL-Interpreter Public

    Real-time Indian Sign Language to text-and-speech interpreter — Flutter app + MediaPipe/TensorFlow inference server, MobileViT for static signs and LSTM for dynamic ones.

    Jupyter Notebook 1 1

  3. SevanaGPT SevanaGPT Public

    Forked from silverdawn-gtm/SevanaGPT

    Multilingual AI assistant for discovering Indian government welfare schemes — chat, hybrid search, and eligibility checks across 4,600+ schemes in 22 languages.

    Python