Aryan · Rudr · Pandit
lat 30.7333° · lng 76.7794°
tok/s 184.2
emb dim 1536
ctx 128k
p(reason) 0.97
attn heads 32
model · llama-3.2-vision
rag · faiss + bm25
agent · react · memory
01001000 01101001
Δt 1.8s · ease(.22,.61,.36,1)
fps 60.0
ARP · v2026.6
aryanrudr911 ↗
AI · Engineer / Creative Technologist
Chitkara · Chandigarh · India

ARYAN

Aryan Rudr Pandit

PANDIT

Scroll · the system unfolds
⊹ 02 / RolesOperating across four disciplines
AI ENGINEER/PRODUCT DESIGNER/MOTION DESIGNER/CREATIVE TECHNOLOGIST/.AI ENGINEER/PRODUCT DESIGNER/MOTION DESIGNER/CREATIVE TECHNOLOGIST/
AI ENGINEER/PRODUCT DESIGNER/MOTION DESIGNER/CREATIVE TECHNOLOGIST/AI ENGINEER/PRODUCT DESIGNER/MOTION DESIGNER/CREATIVE TECHNOLOGIST/
03 / Manifest

Iengineerintelligentsystemsthatreasonwithpurpose,adaptwithcontext,anddeliverwithconfidence.FromautonomousagentstoproductionAI,everyexperienceisdesignedtofeeleffortless,trustworthy,andundeniablyhuman.

+9.25
CGPA · AI/ML
10k+
PDFs indexed
1.1B
params fine-tuned
SIH '24
national finalist
⊹ 04 / Selected Systems

BUILD

2026
Multimodal RAG — Local-First Document Intelligence diagram
fig.01
01 / Work2026

Multimodal RAG

Local-First Document Intelligence

Production pipeline over 10,000+ PDFs. Hybrid BM25 + FAISS retrieval, HyDE query expansion, LLaMA 3.2 Vision for local image captioning, CLIP cross-modal search.

FastAPIFAISSBM25CLIPLLaMA 3.2-VOllama
2025
Confido — AI Interview Analysis
fig.02
02 / Work2025

Confido

AI Interview Analysis

NLP evaluation system parsing interview transcripts into structured feedback. Semantic similarity, scoring logic, Firebase-backed session memory.

PythonNLPFirebaseREST
2024
Empathetic AI — Fine-Tuned LLM + Emotion
fig.03
03 / Work2024

Empathetic AI

Fine-Tuned LLM + Emotion

Fine-tuned TinyLLaMA on conversational datasets. Multimodal inference pipeline integrating emotion detection to dynamically adapt model responses.

TinyLLaMA 1.1BPyTorchTransformersOpenCV
2024
Smart Council — Multi-LLM Reasoning
fig.04
04 / Work2024

Smart Council

Multi-LLM Reasoning

Flask-based multi-agent framework. Prompt-driven argument generation, structured scoring of reasoning quality across model panels.

FlaskLLM APIsPrompt Eng.
⊹ 05 / Stack

THINK

0x0tok7veceemb15ctx1catt230x2atok31vec38emb3fctx46att4d0x54tok5bvec62emb69ctx70att770x7etok85vec8cemb93ctx9aatta10xa8tokafvecb6embbd
LLM SystemsCLUSTER · 01RAGReActTool-UseMemoryHyDEPrompt Eng.Deep LearningCLUSTER · 02PyTorchTransformersCNNsRNNsFine-TuningAgent FrameworksCLUSTER · 03LangChainLangSmithMulti-AgentOrchestrationRetrievalCLUSTER · 04FAISSBM25CLIPsentence-transformersEngineeringCLUSTER · 05FastAPIFlaskDockerPostgreSQLMongoDBNLPCLUSTER · 06spaCyNLTKHuggingFaceNERSentiment
graph · liveidle · scanningfps · 60
node_graph · v1.4streaming · <tokens/s 184>nodes 06 · edges 15
⊹ 06 / Journey

SHIP

2023
Ch. 01

Chitkara

B.Tech · Artificial Intelligence & Machine Learning

Began the formal study of intelligent systems. CGPA 9.25.

2024
Ch. 02

Smart India Hackathon

National-Level Finalist

Built a RAG-powered multi-hazard risk assistant. Selected among top national teams.

2024
Ch. 03

Empathetic AI · Smart Council

Fine-tuning & multi-agent reasoning

Shipped a TinyLLaMA fine-tuned with emotion-aware inference and a Flask multi-LLM evaluation framework.

2025
Ch. 04

Confido

AI-Powered Interview Analysis

NLP evaluation platform — semantic similarity, scoring, structured feedback at scale.

2026
Ch. 05

Multimodal RAG

Local-First Document Intelligence

10,000+ PDFs. Hybrid retrieval. Local vision models. Production-grade ingestion at the edge.

⊹ 07 / Signal
channel · open

— Let's build a system together

aryanrudr911
@gmail.com
Phone
+91 93068 40834
Based
Chandigarh · India
© 2026 Aryan Rudr Pandit · System v2026.6Built · designed · shipped solo