00 Abdul Raheem

Final Year  ·  CBIT Hyderabad  ·  Available for Opportunities

Abdul Raheem.

I build systems that think — across machine learning,
multi-agent AI, and full-stack engineering.

View Projects Download Resume

I build things
that think.

Final-year IT undergrad at CBIT, Hyderabad — CGPA 9.13/10 — with a focus on systems that sit at the intersection of research and production engineering. My work spans multi-agent AI, NLP pipelines, computer vision, and full-stack web platforms — from prototyping to deployment.

I've built autonomous agents that won national hackathons, contributed simulations to India's national ocean science body (INCOIS), and published research at IEEE conferences. I care deeply about the gap between an interesting idea and a working system.

Outside engineering, I've led diplomatic negotiations as a delegate at Model United Nations, directed photography and creative media for cultural events, and designed techno-cultural experiences for student communities. The same systems-thinking applies everywhere.

9.13/10 CGPA
2 IEEE Papers
2026
🏆 HackPrix S3
Winner
INCOIS Project Intern

Where I've worked.

Project Intern

Indian National Centre for Ocean Information Services (INCOIS)  ·  Hyderabad, India

Working on the modernisation of INCOIS's legacy SARAT ocean simulation platform — migrating it from a single-coordinate system into a multi-mode particle seeding framework supporting Multi-Location and Multi-Depth simulation scenarios.

Built a Java-based simulation agent deployed on Apache Tomcat, engineered Bash automation pipelines for remote simulation execution over SSH, and implemented asynchronous Leeway drift modelling with dynamic date-arithmetic-based seeding algorithms to coordinate simulation outputs across geographic and temporal contexts.

Java Apache Tomcat Bash SSH Linux Oceanographic Simulation

Things I've built.

🏆 HackPrix S3 · Best Use of Sarvam AI

AGENTGRID

Autonomous multi-agent energy operating system

An autonomous platform for real-time energy infrastructure management, built for HackPrix S3. Multiple specialised AI agents handle grid monitoring, anomaly detection, and optimisation decisions — orchestrated via LangGraph with natural language interfaces powered by Sarvam AI.

I owned the entire 3D visualisation layer: live energy grid rendering using Three.js and React Three Fiber, with real-time agent state reflected in the 3D scene. Backend powered by FastAPI with full LangGraph multi-agent pipelines.

FastAPI LangGraph Sarvam AI Three.js React Three Fiber Python React
→ GitHub

Meta-LoRA

Few-shot molecular generation via context-conditioned LoRA

A research-grade molecular generation framework combining scaffold-episodic meta-learning with context-conditioned LoRA adapters. Trained on the ZINC250k dataset (224,568 molecules), the system achieves 96.8% SMILES validity and 97.4% uniqueness — outperforming ZINCGene on QED drug-likeness metrics.

Architecture uses 256-d Morgan fingerprint vectors with a dual-branch context encoder, allowing low-rank adaptation of a frozen pre-trained LM without losing chemical grammar constraints. A three-phase training pipeline with a 4-layer Pre-LN causal Transformer (3.19M parameters) freezes all weights post pre-training to preserve chemical grammar through adaptation.

PyTorch Transformers LoRA RDKit SMILES Meta-Learning Python
→ GitHub

ApplicationTracker

Intelligent job application management, end-to-end

A full-stack platform that removes the chaos from job hunting. An NLP extraction engine built on spaCy parses company names, roles, and application statuses directly from raw email text — no manual entry required.

Features async job sync via Celery, a PostgreSQL + Redis backend, JWT authentication with encrypted credential storage, and a real-time search dashboard for filtering and tracking applications at scale. A deterministic NLP pipeline using Regex and sender-domain heuristics handles edge cases where structured data is absent.

FastAPI Next.js PostgreSQL Redis Celery spaCy JWT Python
→ GitHub

ATLAS-GAME

Graph-theoretic analysis of a turn-based geography game

Models the children's word-chain geography game Atlas as a directed graph — where nodes are countries and cities, and edges connect places whose last letter matches the next place's first letter. The objective: uncover winning strategies using network science.

The analysis applies centrality measures to mathematically define "winning" positions, detects community structures within the game graph, and trains a Graph Neural Network for link prediction. Built entirely in a single Jupyter notebook environment (Colab-compatible), processing a cleaned dataset of world countries and cities.

Python NetworkX PyTorch PyTorch Geometric Graph Neural Networks Pandas Matplotlib
→ GitHub

Published work.

Evaluating LLM Reasoning on String Rewriting Puzzles:
Prompting Strategies, Metrics, and Human–AI Comparison

IEEE AI-SIS 2026

An empirical study on how well large language models handle formal string rewriting tasks — benchmarking prompting strategies, proposing custom evaluation metrics, and comparing LLM performance directly against humans.

→ View Paper

Lightweight CNN Outperforms ImageNet-Pretrained ResNet18
for Multi-Class Alzheimer's Classification: A Domain Mismatch Analysis

IEEE ICSCSS 2026

Shows that a purpose-built lightweight CNN can outperform transfer-learned ResNet18 on Alzheimer's MRI classification — questioning the assumption that pretrained ImageNet models are always the better starting point for medical imaging tasks.

→ View Paper

What I work with.

LANGUAGES
Python Java C SQL JavaScript HTML/CSS Python Java C SQL JavaScript HTML/CSS
FRAMEWORKS
PyTorch TensorFlow Keras scikit-learn FastAPI React.js Node.js Express.js Pandas NumPy NetworkX Kafka PyTorch TensorFlow Keras scikit-learn FastAPI React.js Node.js Express.js Pandas NumPy NetworkX Kafka
DATABASES
PostgreSQL MongoDB MySQL Oracle Hadoop HiveQL Spark SQL PostgreSQL MongoDB MySQL Oracle Hadoop HiveQL Spark SQL
AI / ML
AI Agents NLP Computer Vision Deep Learning LLMs RAG Pipelines Multi-Agent Systems AI Agents NLP Computer Vision Deep Learning LLMs RAG Pipelines Multi-Agent Systems
TOOLS
Git GitHub GitLab Postman Linux/Unix Cursor Docker Git GitHub GitLab Postman Linux/Unix Cursor Docker

Validated expertise.

M

MongoDB Associate Developer

Python · MongoDB University

Validates proficiency in building Python applications with MongoDB, including schema design, indexing, and aggregation pipelines.

View Certificate →
S

Certified Agentforce Specialist

Salesforce

Validates ability to build, configure, and deploy AI agents using Salesforce's Agentforce platform and ecosystem.

View Certificate →
O

Oracle Certified Foundations Associate

Oracle University

Validates foundational knowledge of Oracle Cloud Infrastructure, cloud concepts, and core OCI services.

View Certificate →

Recognition.

Competitive & Academic

  • HackPrix S3 · Best Use of Sarvam AI Track National-level AI hackathon · 2025
  • Special Mention · COSC Hacktoberfest 2025
  • Amazon ML Summer School 2025 Competitive national selection programme by Amazon
  • ALMA Maths Olympiad 2018 · Silver Medal

Leadership & Recognition

  • High Commendation · VASAVI MUN 2024 (UNHRC)
  • High Commendation · GNITS MUN 2025 (UNHRC)
  • Excellence Award · CBITMUN MUN 2025 (UNCSW)
  • Delegate · GNITS MUN 2026
  • Chief Advisor · VMedha CBIT Led 45-member team organising annual techno-cultural fest · August 2025 – March 2026

Have something
interesting?

Let's talk.

Open to research collaborations, AI/ML roles, and ambitious projects.