Skip to content
View shubhamgupta407's full-sized avatar
🎯
Focusing
🎯
Focusing
4 followers · 0 following

Block or report shubhamgupta407

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. 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
shubhamgupta407/README.md

Shubham Gupta

Agentic AI · Trustworthy LLMs · World Models

Most AI systems generate. The interesting ones reason, retrieve, adapt, and decide.

🧠 Who I am

B.Tech CSE (AI & ML) @ LPU · Project Assistant @ IIT Kharagpur (prev. Research Intern) under Dr. Ritesh Sarkhel (ex-Amazon).

I work on the part of AI that isn't just training - reasoning, coordination, and deciding what to do next.

🎯 What drives me

When does a multi-agent system actually need another agent?

And the follow-up that keeps me up:

How do we make AI reliable when the answer actually matters?

🔬 Research

IIT Kharagpur - Project Assistant (prev. Research Intern)

⟶ Agent Spawning · with IBM Graph-theoretic + information-theoretic approaches to when/where a multi-agent system should spawn agents. Reproducing and analysing multi-agent collaboration benchmarks.

⟶ Zero-Day Fraud Detection · with Persistent Systems Contrastive representation learning · hard-negative mining · temporal generalisation · world-model-based sequence modelling.

ROC-AUC: 0.70 → 0.84 on zero-day fraud

⟶ World Models & Agentic AI How agents reason about outcomes before acting. JEPA DreamerV3 TransDreamer STORM IRIS WALL-E 2.0 WorldCoder

IIT (BHU) - Research Collaboration Credibility-aware RAG · trustworthy NLP · reliable LLM systems.

💼 Experience

Role Where When
Project Assistant (prev. Research Intern) IIT Kharagpur 2026 - Present
Machine Learning Intern Unified Mentor Pvt. Ltd. Nov 2025 - Feb 2026

🚀 Projects
Sentinels of Truth
Multi-agent fact verification		 LangGraph LLMs Tavily FastAPI React
SHL Intelligence Console
Conversational assessment recommender		 LangGraph FAISS MiniLM FastAPI
Sales Call Intelligence
RAG-based sales analytics		 RAG FAISS LangChain LLMs
ASL Detector
Real-time sign language CV		 TensorFlow OpenCV CNN
Job Role Clustering
Unsupervised role discovery		 Sentence Transformers Scikit-learn

🏆 Highlights

TCS Prime Offer · ₹10 LPA   NPTEL HCI · 94% Gold   250+ DSA   Research @ IIT KGP   IBM   Persistent Systems   IIT (BHU)

🛠️ Stack

📈 Stats

Building. Breaking. Making AI actually useful.

🤝 Let's Connect Working on something in agentic systems, trustworthy AI, or applied ML? I'd genuinely like to hear about it.

Still building. Still breaking things. Still figuring out better ways to make AI actually useful.

Pinned Loading

  1. ksolves-resolveai ksolves-resolveai Public

    Hybrid enterprise AI support automation platform using local models, semantic routing, LLM fallback, audit logs, and human escalations.

    JavaScript 1

  2. sales-call-intelligence-system sales-call-intelligence-system Public

    AI-powered enterprise SaaS — sales call intelligence platform with RAG pipeline, FAISS vector search, real-time analytics, system topology, audit logs, and intelligence deck generation

    JavaScript 1

  3. Gitanalyze Gitanalyze Public

    Built for technical hiring teams who are tired of opening 10 GitHub tabs per candidate. GitAnalyze automates the entire profile evaluation process — scoring, stack detection, project analysis, and …

    JavaScript 2

  4. Serverless-Event-Platform Serverless-Event-Platform Public

    Serverless event management platform with real-time updates, built using React and Firebase, featuring authentication, dashboards, and event workflows.

    JavaScript 1

  5. Job-role-Clustering-using-unsupervised-learning Job-role-Clustering-using-unsupervised-learning Public

    Built an unsupervised NLP system to cluster LinkedIn-style job descriptions using Sentence-BERT embeddings and KMeans. Performed text cleaning with NLTK, generated semantic embeddings, evaluated cl…

    HTML 1

  6. LLM-Reliability-Verification-Patent LLM-Reliability-Verification-Patent Public

    A patented AI system that gives every LLM output a quantified trust score using independent verification + multi-pass inference consistency analysis.

    1