solitarybear.

AMAN AGARWAL / BHOPAL, INDIA

From data
to intelligence.

Exploring patterns. Training models.
Building useful software with AI and machine learning.

ECE undergraduate at IIIT Bhopal, focused on deep learning, computer vision, and data-driven applications.

Explore my work GitHub
FIELD NOTES / 001AI + MACHINE LEARNING
From data to learned representations An illustrative neural network connects input data through hidden layers to predictions. This is a conceptual diagram, not a specific model architecture. DATAREPRESENTATIONSPREDICT LEARN THE PATTERN.BUILD THE POSSIBILITY.
A question I keep coming back to:
How can data become useful intelligence?
DEEP LEARNINGMACHINE LEARNINGCOMPUTER VISIONPYTHON

01 / SELECTED WORK

Ideas, put into practice.

All repositories

01 — MULTIMODAL DEEP LEARNING

Learning across
vision and language.

SemGrasp combines visual input and natural-language task descriptions to predict grasp quality. A robotics problem becomes a study in multimodal representation learning.

My work compares a cross-modal transformer against a FiLM-conditioned ResNet baseline, exploring how a physics-informed head supports task-conditioned prediction.

  • PyTorch
  • Transformers
  • Computer vision
Explore SemGrasp

TWO MODALITIES. ONE PREDICTION.

Visual and task inputs combine in a cross-modal transformer to produce a grasp score RGB-D INPUTvisual featuresTASK TEXTlanguage featuresCROSS-MODAL TRANSFORMERGRASP SCORE

Concept illustration · Multimodal prediction

02 — DATA ANALYTICS / PYTHON

Daily Score Tracker

An offline Python app that makes everyday activity data useful. Its analytics layer explores trends, completion rates, and cross-activity correlations, with versioned scoring rules that preserve historical records.

  • Python
  • GTK4 / libadwaita
  • SQLite
View project

03 — PYTHON / DATA SYSTEMS

PocketLog

An offline expense-sharing app for shared costs and clear settlements. Equal or custom splits, participant reports, and integer-paise arithmetic keep the numbers precise.

  • Python
  • PySide6 / Qt
  • SQLite
View project

02 / A LITTLE CONTEXT

Curious about data.
Driven to build.

I’m Aman, also known online as Solitarybear. I’m pursuing a B.Tech in Electronics and Communication Engineering at IIIT Bhopal, with a focus on artificial intelligence, machine learning, and the software that makes them useful.

My work spans multimodal deep learning, computer vision, and data-focused Python applications. I’m interested in the full path from preparing data and evaluating models to building tools people can use.

Find me on LinkedIn

MAY — JUL 2025

ML Research Intern

IIT (BHU), Varanasi

Developed ANN and recurrent-network approaches for joint-angle prediction, combined sensor streams for real-time estimation, and worked with PPO/SAC reinforcement-learning control. Applied this ML research to humanoid-robot stability.

  • Neural networks
  • Sensor fusion
  • Reinforcement learning

SEP 2023 — JUN 2027 (EXPECTED)

B.Tech · ECE

IIIT Bhopal

Coursework in machine learning, neural networks, probability, linear algebra, and data structures.

NOV 2025 — PRESENT

Building with a team

Spark Robotics Club · IIIT Bhopal

Core technical team member contributing to competition robots and technical workshops. Runner-up in the 2024 college internal hackathon with a precision-agriculture project combining drone simulation and ML-based yield prediction.

03 / TOOLKIT

Tools behind the work.

/ 01

Machine learning & vision

PyTorch · TensorFlow · scikit-learn · Hugging Face Transformers · OpenCV · Neural networks

/ 02

GenAI & data

RAG · LangChain / LangGraph · ChromaDB / Pinecone · NumPy · Pandas · SQL · ETL pipelines

/ 03

Software & deployment

Python · C++ · FastAPI · Flask · Docker · Git · Linux · Azure · Databricks · ONNX

04 / LET’S CONNECT

Have something
interesting in mind?

↗

For conversations about AI, machine learning research, and useful software.

GitHub ↗LinkedIn ↗