SHASHWAT // SYSTEM ARCHIVE

∞ ── TECH.STACK

Stack

Tools and technologies I use to build ML systems, agentic pipelines, and full-stack applications. Focused on the intersection of research and production.

Languages

  • Python

    Primary language for all ML/AI work

  • C++

    Performance-critical ML components

  • TypeScript

    Frontend & fullstack applications

  • SQL

    Data wrangling and analytics

  • Bash

    Scripting, automation, server ops

ML / Deep Learning

  • PyTorch

    Primary DL framework for all model work

  • HuggingFace Transformers

    Fine-tuning and inference of LLMs

  • scikit-learn

    Classical ML, preprocessing, evaluation

  • NumPy / Pandas

    Data manipulation and numerical ops

  • OpenCV

    Computer vision pipelines

  • Weights & Biases

    Experiment tracking and visualisation

LLMs & Agents

  • LangChain / LangGraph

    Agentic pipelines and state machines

  • OpenAI API

    GPT-4 and embedding API integration

  • Groq (Llama 3)

    Ultra-fast inference for agent tasks

  • Ollama

    Local model serving and prototyping

  • Pinecone / ChromaDB

    Vector stores for RAG systems

  • Streamlit / Gradio

    Rapid ML app prototyping

Backend & APIs

  • FastAPI

    Async REST APIs for ML model serving

  • Next.js

    Full-stack web apps with server components

  • PostgreSQL

    Primary relational database

  • Redis

    Caching, queues, session management

  • Docker

    Containerised model deployment

Tools & Infra

  • Git / GitHub

    Version control, CI/CD via Actions

  • Jupyter / VS Code

    Exploration and production code

  • Linux (Ubuntu)

    Primary dev and server environment

  • Vercel

    Frontend deployment and edge functions

  • Cursor

    AI-assisted development

Currently Learning

  • Triton

    Custom CUDA kernels for GPU ops

  • CUDA / C++ extensions

    Low-level DL performance engineering

  • Rust

    Systems programming for ML tooling