Skip to content
View shashank-518's full-sized avatar

Block or report shashank-518

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.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. 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
shashank-518/README.md

πŸ‘‹ Hi, I'm Shashank S!

Typing SVG

coding animation

profile views


πŸ’« About Me

  • πŸ€– AI Engineer, founding team at Digitomics AI (Bengaluru). I build a production GenAI platform for cloud, AI and SaaS spend (FinOps).
  • 🧠 I work on LLM agents, RAG pipelines, model routing, tool calling and realtime voice AI.
  • ⚑ Brought chatbot response latency down from ~3s to ~800ms with task-aware model routing.
  • πŸ“š Built a multi-tenant RAG pipeline (tenant-aware knowledge bases) and 2 production agents (FinOps and DevOps).
  • πŸ‘¨β€πŸ« Led and mentored 12 interns through architecture discussions, code reviews and feature ownership.
  • πŸ† Hackathon 2025: Top 20 of 1,000 with RakshaMitra, a public-safety app.
  • πŸŽ“ B.Tech in Computer Science and Engineering, Presidency University (2022–2026).
  • 🌱 Currently exploring: LLM evaluation (Ragas, DeepEval), fine-tuning with LoRA/PEFT, and MCP.

πŸš€ Featured Projects

Project What it does Stack
πŸŽ™οΈ AI Voice Calendar Agent Multi-user voice agent that schedules meetings and reminders by voice. WebRTC realtime audio with barge-in, per-user tool calling, Google OAuth + Calendar. Azure OpenAI Realtime, FastAPI, React, PostgreSQL, Docker
πŸ–ΌοΈ Multimodal Local RAG Chat with PDFs, images and videos (OCR + Whisper). Fully local, with citations by page and video timestamp. Ollama (Mistral), ChromaDB, Whisper, Tesseract, Streamlit
πŸ“„ RAG Document Q&A Grounded document Q&A: 92% context accuracy over 500+ queries, 40% fewer hallucinations, sub-2s responses. LangChain, OpenAI, ChromaDB
🏸 Shuttle: Domain LLM Fine-Tuning LoRA fine-tuning of SmolLM2-360M on a badminton Q&A set, training ~1.2% of parameters, with a chat UI. PyTorch, Hugging Face, PEFT/LoRA, FastAPI
🌐 Fact-Aware Cross-Lingual Summarization Research project: a fact-aware training objective on mBART for Hindiβ†’English news summarization. PyTorch, mBART, ROUGE, BERTScore
πŸ“ˆ State-Wise Sales Forecasting Trains ARIMA, SARIMA, Prophet, XGBoost and LSTM per US state and picks the best model automatically. statsmodels, Prophet, XGBoost, Keras, FastAPI, Docker
πŸ—£οΈ Voice RAG Support Assistant E-commerce support bot: speech-to-text β†’ RAG β†’ LLM β†’ text-to-speech. Whisper, Gemini, ChromaDB, FastAPI
🏠 EstateAI Conversational property search driven by strict JSON structured outputs. Gemini, FastAPI, Pydantic, React

πŸ”— Portfolio: shashank-mu.vercel.app


🌐 Socials

LinkedIn Portfolio Email


πŸ’» Tech Stack

Languages: Python TypeScript JavaScript SQL Java

AI / GenAI: OpenAI Azure OpenAI Gemini Ollama Groq PyTorch Hugging Face LoRA/PEFT Whisper

RAG / Agents: LangChain LangGraph LlamaIndex ChromaDB FAISS pgvector MCP

LLM Evaluation: Ragas DeepEval LangSmith Langfuse

ML / Data: scikit-learn TensorFlow Keras XGBoost Pandas NumPy

Backend: FastAPI NodeJS Express.js Flask Pydantic GraphQL Kafka JWT

Frontend: React Next JS Redux Tailwind CSS Vite Streamlit

Databases: PostgreSQL MongoDB MySQL Redis Elasticsearch Prisma

Cloud / MLOps: Azure AWS Docker Kubernetes GitHub Actions MLflow Vercel

AI Dev Tools: Claude Code Cursor OpenAI Codex Git Postman Linux


πŸ“œ Certifications

  • 🧩 Introduction to Model Context Protocol: Anthropic (2026)
  • 🧠 Data Science, Machine Learning, Deep Learning & NLP: Udemy (2025)
  • 🌐 React, Node.js, Express & MongoDB (MERN): Udemy (2025)

πŸ“Š GitHub Stats

GitHub Stats GitHub Streak Most Used Languages

πŸ† GitHub Trophies


✍️ Random Dev Quote


Building AI systems that work in production. πŸš€

Pinned Loading

  1. Forecasting Forecasting Public

    A production-grade machine learning forecasting system that predicts the next 8 weeks of beverage sales for each US state. The system automatically trains multiple time-series models (ARIMA, SARIMA…

    Jupyter Notebook