Skip to content

AI Engineer Intern at Estrel.ai

Satyam Raj

I turn complex workflows into intelligent, AI-powered products.

AI Engineer Intern building practical AI systems, from LLM-powered applications to automated workflows.

About

I'm Satyam Raj, a final-year B.Tech student in Computer Science & Engineering (AI & ML) and an AI Engineer Intern at Estrel.ai, where I get to ship AI software rather than just study it.

Most of my work centers on one practical problem: making complex information readable and repetitive work automatic. That has meant documentation automation tools, retrieval-augmented generation (RAG) pipelines, and applications that let people ask questions of their own data instead of digging through files.

Working across a final-year degree and a real internship has taught me to treat AI as an engineering problem, not a demo: systems count only when they work outside a notebook.

I build with Python, FastAPI, and LangChain, rounded out by the web stack needed to take a model from idea to interface.

Experience

  1. Generative AI Intern

    Jan 2025 to Mar 2025, Remote

    EduSkills Foundation x Google Cloud

    RAG on Vertex AI and Gemini

    • Built GenAI applications with Vertex AI, Gemini APIs, RAG, prompt engineering, vector search, and structured outputs.
  2. Agentic AI Intern (Capstone)

    Dec 2025 to Jan 2026, Remote

    Wiiz Platform

    Best Capstone Project

    • Built a multi-agent Contract Risk Analyzer with LangChain to extract, classify, and summarize risks from legal contracts.
    • Won Best Capstone Project and Highest Workflow Creation for the program.
  3. AI Engineer Intern

    May 2026 to Present, Remote

    Estrel.ai

    60% less manual paperwork

    • Engineer AI workflow automation with Python, LLMs, and OCR for enterprise business processes.
    • Built document-processing applications that automated paperwork and reduced manual effort by 60% across internal workflows.

Projects

Smart India Hackathon 2026 prototype for a Ministry of Coal problem statement

  • Turns mining and geological PDFs, scans, images, CSVs, and spreadsheets into structured, traceable facts.
  • Grounded Q&A with page citations, automated report generation with fact checks, topic analysis, and a human review queue for cross-source conflicts.
  • 109 passing offline tests. Deployed on Vercel.
  • FastAPI
  • Next.js 14
  • PostgreSQL
  • pgvector
  • RapidOCR
  • pdfplumber
  • RAG
  • Tailwind
  • Docker Compose
  • Groq LLM

GitHubLive site

One chat interface for every model, plus a drop-in OpenAI-compatible gateway

  • Self-hosted, provider-agnostic LLM playground on FastAPI, LangChain, and LangGraph. Point it at any provider, model, base URL, or key.
  • Live token streaming and per-thread memory over a single interface.
  • Exposes an OpenAI-compatible `/v1` endpoint, so existing tools can be pointed at it without changes.
  • Python
  • FastAPI
  • LangChain
  • LangGraph

GitHub

Machine learning that predicts a machine's operating condition from live sensor readings

  • Predicts operating condition from sensor data and production KPIs to surface early warning signs of failure, downtime, and quality loss.
  • Targets the cases where fixed threshold rules fall short, which is where a rules-based alert is least useful.
  • Python
  • Machine Learning
  • Sensor Analytics

GitHub

An interactive machine learning lab where concepts can be visualised, calculated, and broken

  • Each concept is worked through on one page: understood, visualised, calculated, built, experimented on, broken, and practised.
  • Built as a static, fast-loading teaching surface rather than a slide deck.
  • Next.js App Router
  • React 19
  • TypeScript
  • Tailwind CSS v4
  • SSG

GitHub

Agentic task management that turns a natural-language goal into tracked subtasks

  • Understands natural-language goals, decomposes them into subtasks, and processes them with an LLM.
  • Persistent contextual memory backed by RAG, so tasks carry context across sessions.
  • Full-stack with JWT authentication and password reset. Deployed.
  • Next.js
  • LangChain
  • Supabase
  • PostgreSQL
  • RAG
  • Vercel
  • Render

GitHubLive site

Retrieval augmented analysis over a CSV dataset

  • Upload a CSV dataset, query it in natural language, and get context-aware insights.
  • Forks from readers who wanted to explore the same dataset, which is the clearest signal that it solved a real problem.
  • Python
  • LangChain
  • FastAPI
  • Pandas

GitHub

A minimalist sanctuary for word exploration

  • Editorial-grade dictionary experience with mesh gradients, noise overlays, and considered typography.
  • Smart typeahead for word lookup, built for reading rather than for speed.
  • Next.js 16
  • TypeScript
  • Tailwind CSS v4
  • Framer Motion

GitHubLive site

An implementation-first ML and AI engineering learning system

  • A deeply connected curriculum running from fundamentals through deep learning, NLP, generative AI, and LLMs.
  • Continues into RAG, LangChain and LangGraph, AI agents, evaluation, and deployment.
  • Built as connected modules and projects rather than isolated notebooks.
  • Python
  • Jupyter
  • Machine Learning
  • Deep Learning
  • NLP
  • RAG

GitHub

Skills

Grouped by what they are used for, with AI and machine learning first because that is the work. No self-assessed levels, since a percentage like “Python 90%” is unmeasurable and means nothing to a reader. The tools the production work runs on are the ones drawn at full contrast in the first group.

AI and machine learning

  • LangChain
  • LangGraph
  • RAG
  • Hugging Face
  • Scikit-learn
  • Machine Learning
  • Deep Learning
  • NLP
  • Generative AI

AI models and APIs

  • Groq
  • OpenAI API
  • Google Gemini
  • OpenRouter

Frontend

  • React
  • Next.js
  • Framer Motion
  • Tailwind CSS
  • shadcn/ui
  • HTML
  • CSS

Backend

  • FastAPI
  • Node.js
  • Express.js
  • REST APIs

Data and vector stores

  • PostgreSQL
  • pgvector
  • Supabase
  • ChromaDB
  • MongoDB
  • SQLite
  • Pandas

Programming languages

  • Python
  • TypeScript
  • JavaScript
  • Java
  • SQL

Tooling and deployment

  • Docker
  • Git
  • GitHub
  • Vercel

Achievements and Certifications

  • Top 100 Finalist, Agentic AI Buildathon 2026

    2026

    Team award. Team Pentos, Capgemini Exceller

  • Best Capstone Project, Enterprise Contract Intelligence and Risk Analyzer

    Team award. WiiZ x ESTREL.AI Agentic AI Program

  • Certificate Program in Artificial Intelligence and Machine Learning, Gold (70% and above)

    Aligned to IT-ITeS Sector Skills Council, NASSCOM competency standards

  • Machine Learning and Pattern Recognition, Deep Learning with Python, TensorFlow and Keras, and Neural Networks and Deep Learning

    Dec 2025

    Samatrix Consulting Private Limited

  • R Programming, Data Analysis using Python, and Probability Modelling using Python

    Apr 2025

    Samatrix Consulting Private Limited

  • Introduction to Modern AI

    Jan 2025

    Cisco

  • Free Java Certification Course

    Nov 2024

    DataFlair

  • Certificate of Participation, Agentic AI Program

    Team award. WiiZ x ESTREL.AI

Contact

Open to full-time AI engineering opportunities starting in 2027.