3D NEURAL SOLAR CONSTELLATION

The Core Nucleus.

Hover over any orbiting satellite node to inspect my educational history, AI neural model training background, and core software engineering pillars.

ACADEMIC HISTORY SATELLITES

KIIT University

B.Tech CSE

B.Tech in CSE (AI & ML Specialization)

Specializing in PyTorch Deep Neural Architectures, Computer Vision, and High-Performance ML Systems.

Bhubaneswar, Odisha, India2023 to 2027

Notre Dame College

GPA 5.00

Higher Secondary Certificate (HSC)

Completed HSC with distinction and top merit in Science & Computer Science stream.

Dhaka, Bangladesh2022

St. Gregory’s High School

GPA 5.00

Secondary School Certificate (SSC)

Completed SSC with perfect GPA 5.00 in Science division with strong mathematical foundation.

Dhaka, Bangladesh2020

NEURAL CORE NUCLEUS

DHAKA, BANGLADESH

Dabojit

DABOJIT SARKAR

CSE (AI & ML) ASPIRANT & FULL STACK DEV

Download Resume GitHub Profile Sign Wall

AI ENGINEERING SATELLITES

AI & ML Core Specialization

PyTorch & Scikit-Learn

Deep Learning & Neural Architectures

Designing custom 3-block CNN acoustic models, 40×300 MFCC matrix classifiers, and edge CPU inference pipelines.

Research & EngineeringCore Focus

Generative AI & LLM Systems

Gemini + Groq + RAG

Multi-Model Grounding Engines

Architecting dual-LLM document risk analysis tools with verbatim paragraph offset citations and zero hallucinations.

NLP Offset CitationCore Focus

Full-Stack & WebGL Engineering

Next.js 15 & Three.js

Tactile Interactive UIs

Building high-performance reactive web platforms with 3D WebGL graphics, Framer Motion, and REST/MLOps backends.

Real-Time TelemetryCore Focus

INTERACTIVE AI PIPELINE ARCHITECTURES

Neural Pipeline Execution & Flow Simulator

SoundProof AI Audio Pipeline

PyTorch 3-Block CNN Classifier

93% ASVspoof 2019<1.0s on CPU

01. Raw Audio WAV

Audio File Ingestion & Librosa Loader

02. 40×300 MFCC Matrix

Acoustic Fingerprint Spectrogram Conversion

03. 3-Block PyTorch CNN

600K Parameter Conv2D Feature Extraction

04. Real / Fake Classification

Confidence Score Output (Human vs Synthetic)

MODEL TELEMETRY MATRIX

CUDA ACTIVE

Training Loss

0.0412 ↓

Validation Acc

98.4% ↑

SELECT ARCHITECTURE MODEL

SoundProof Audio ML Engine

40×300 MFCC Matrix • 600K Parameter CNN • <1s CPU Latency • ASVspoof 2019

AI SKILL TELEMETRY RADAR

LIVE METRICS

Deep Neural Architectures (PyTorch / TF)98%

Generative AI & LLMs (Gemini / Groq)96%

Full-Stack WebGL UIs (Next.js 15 / Three.js)95%

Production MLOps & REST (FastAPI / Docker)92%

<1s CPU Inference

93% ASVspoof Acc

Zero Hallucinations

Leave a mark on my wall

Sign my digital guestbook with a message, feedback, or hello.

Go to Guestbook