Open to internships & roles

Final-year CS · AI & ML · Agentic Systems

I buildagentic systemsthat ship.

Shouryaman Purohit · final-year CS at VIT Bhopal. Multi-agent pipelines, RAG, and products from prototype to deployment.

CrewAIOpenAI Agents SDKVertex AIFine-tuningPythonFastAPI

CGPA

8.94

VIT Bhopal

LeetCode

750+

Knight tier

IEEE

Author

PIPO · Global AI Summit 2025

Hackathon

Top 10

Google Crowdsource Adversarial Nibbler

Scroll

About

Intelligent systems, built to ship.

Final-year CS student at VIT Bhopal specializing in AI & ML — focused on agentic pipelines, RAG, and products people can actually use.

I work at the intersection of multi-agent AI, retrieval systems, and full-stack delivery.

From LangChain and CrewAI pipelines to MCP-powered tools and cloud-deployed apps — recent work includes Vertex AI fine-tuning at SmartBridge, leading Toastmasters at VIT Bhopal, and shipping Curiosity, Medically, Connected, and HireFlow.

Shouryaman Purohit — portrait

Shouryaman Purohit

Final-year CS · AI & ML · Agentic Systems

Focus

Agentic AI & RAG

Education

VIT Bhopal · AI & ML

Location

Bhopal, India

Now

Building HireFlow

Work

Case studies, not just thumbnails.

Each project follows a clear arc — problem, approach, and impact — with live demos where available.

Shipped & live

Flagship projects deployed and ready to explore.

Curiosity

My role · Built the CrewAI pipeline, FastAPI backend, and React deployment stack.

Live

Multi-agent engineering platform — natural language in, design docs, Python modules, and Gradio demos out, with live SSE streaming in the UI.

Problem

Turning vague product ideas into working Python demos took too many manual handoffs.

Approach

CrewAI engineering crew (design, backend, Gradio UI) with FastAPI + SSE and a React/Vite frontend — deployed on Render.

Impact

Live at curosity.onrender.com — one URL for the full builder and artifact downloads.

CrewAIFastAPIReactGradioSSE

Connected

My role · Full-stack owner — API, frontend, and AWS deployment.

Live

Full-stack developer networking platform for showcasing projects, reviews, and conversations between builders.

Problem

Students and builders lacked a focused space to discover peers and opportunities.

Approach

Django REST Framework backend, JWT auth, PostgreSQL on AWS RDS, and S3 for media — deployed on Render.

Impact

Used by 30+ developers with secure auth, email workflows, and 10+ REST endpoints.

Django RESTPostgreSQLAWSJWTRender

Medically

My role · Designed the RAG pipeline, FAISS retrieval layer, and Streamlit UI.

Live

AI medical assistant built on a RAG pipeline — grounded answers from your knowledge base via semantic search and Gemini.

Problem

Generic LLM answers lacked grounded medical context and citation-style retrieval.

Approach

HuggingFace embeddings + FAISS vector store, LangChain orchestration, and Gemini for generation — Streamlit chat UI.

Impact

Live Streamlit demo with PDF ingestion and context-rich medical Q&A.

RAGGeminiFAISSStreamlitHuggingFace

More work

Additional builds and experiments.

Portfolio AI Assistant

My role · Designed and built the assistant, knowledge base, and API layer.

This site's floating assistant — RAG knowledge base, FAISS search, and OpenAI-powered replies.

You're looking at it right now — hit Ask AI in the corner. That's the demo.

Problem

Visitors needed a way to ask questions without reading every section.

Approach

FastAPI backend, FAISS retrieval, and a Next.js chat widget on this portfolio.

Impact

Always-on virtual rep with memory for unknown questions via admin workflow.

FastAPIOpenAIFAISSNext.js

Deep Research Sidekick

My role · Solo builder — agent orchestration, Serper search, Pushover delivery, and Hugging Face deployment.

Live

Multi-agent deep research assistant (Challenger Deep 2.0) — plan a topic, search the web, and ship a structured report.

Problem

Deep research means planning queries, searching, synthesizing sources, and formatting a report — all by hand.

Approach

OpenAI Agents SDK with dedicated agents for planning, Serper-powered search, report writing, and orchestration — live on Hugging Face Spaces with finished reports pushed to phone via Pushover.

Impact

Live demo on Hugging Face — enter a research topic and get a multi-step report in the app and on your phone.

OpenAI Agents SDKSerperMulti-agentHugging FacePushover

StockVerse

My role · Solo builder — CrewAI agent teams, analysis pipeline, notifications, and Hugging Face deployment.

Live

CrewAI market research crew — pick a domain or use a preset, and get a synthesized market report in the app and on your phone.

Problem

Market research across a domain means coordinating searches, cross-checking findings, and writing a clear report — slow to do manually.

Approach

CrewAI SDK with a research team for the user’s domain, an analyst agent to synthesize team output, and a report writer — deployed on Hugging Face Spaces with results displayed in-app and pushed via notification.

Impact

Live on Hugging Face — specify a market domain and receive a comprehensive report without hand-stitching research.

CrewAIMulti-agentMarket researchHugging FacePushover

PlantDoc

My role · Built the TensorFlow CNN, training pipeline, and Streamlit dashboard.

Plant disease recognition from a leaf photo — CNN-powered diagnosis across 38 crop and disease categories.

Problem

Farmers and gardeners need fast disease ID without waiting on an agronomist.

Approach

Custom Keras CNN trained on ~87K RGB leaf images, wrapped in a Streamlit app — upload, preview, predict.

Impact

Near-instant predictions from apple scab to tomato leaf curl in a simple web UI.

TensorFlowCNNStreamlitComputer VisionPython

That's the highlight reel — the full catalog of experiments lives on my GitHub. Go browse; I don't gatekeep good builds.

Building now

Active development — ask my AI for the latest progress.

HireFlow

My role · Solo builder — agents, MCP integrations, and product UX.

In development

Agentic web app: paste a resume and job URL — AI agents analyze the role, rewrite the resume, and research the company via MCP tools.

Problem

Tailoring resumes and researching companies for each role is slow and repetitive.

Approach

OpenAI Agents SDK + MCP tools for company research and resume rewriting.

Impact

In active development — ask my AI assistant for the latest progress.

MCPOpenAI AgentsMulti-agentFastAPI

Experience

Where I've applied the work.

Internships, leadership, and hands-on delivery.

Period

2025

SmartBridge

Generative AI Intern

  • Fine-tuned models on Vertex AI and shipped generative AI solutions end-to-end.
  • Worked across the full pipeline from experimentation to deployment-ready deliverables.

Period

2024 – Present

Toastmasters Club, VIT Bhopal

President

  • Lead club operations, meetings, and member development programs.
  • Built strong communication and leadership skills alongside technical work.

Toolkit

Tools I reach for.

Agentic AI, full-stack delivery, and cloud — backed by Vertex AI internship work, shipped projects, and LeetCode Knight.

01 · Agentic AI

Agentic AI

  • CrewAI
  • OpenAI Agents SDK
  • MCP
  • LangChain
  • RAG
  • FAISS
  • LiteLLM
  • Gemini

02 · ML & GenAI

ML & GenAI

  • Vertex AI
  • Fine-tuning
  • Prompt Engineering
  • HuggingFace
  • TensorFlow
  • Keras
  • Streamlit
  • OpenCV

03 · Full-Stack

Full-Stack

  • Python
  • FastAPI
  • Django REST
  • React
  • TypeScript
  • Next.js
  • PostgreSQL
  • Docker
  • Redis

04 · Ship & Proof

Ship & Proof

  • AWS RDS · S3
  • Vercel · Railway
  • LeetCode Knight
  • 750+ DSA
  • IEEE Author · PIPO
  • Google Cloud GenAI
  • CodeChef 3★
  • Toastmasters

Contact

Let's build something.

Recruiting, collaborating, or curious — reach out directly or ask my AI assistant.

AI assistant

Ask about my work, skills, or how we can collaborate.