Software Engineer | Full Stack Developer | AI Systems Engineer

Pavankumar S
Kallur

Engineering order from chaos, finding simplicity in complex data. Driven by the philosophy of finding beauty in raw, imperfect, and evolving systems.

Pavankumar S Kallur
"True depth in intelligence lies not in flawless execution, but in how we handle the noise, the outliers, and the beautiful imperfections of the real world."

About My Craft

I am a Computer Science graduate specializing in Artificial Intelligence and Machine Learning, with hands-on experience in deep learning, computer vision, and full-stack software development. My approach combines rigorous mathematics and core CS foundations with a designer's eye for clean, reliable system architecture.

Just as the *wabi-sabi* philosophy appreciates worn surfaces and asymmetric lines, I build AI systems that embrace noisy, real-world logs, shifting user patterns, and unstructured text documents. I thrive on translating chaotic datasets into predictive models and automated pipelines that deliver tangible insights.

CHRONICLES OF CREATION — CHRONICLES OF CREATION — CHRONICLES OF CREATION — CHRONICLES OF CREATION — CHRONICLES OF CREATION — CHRONICLES OF CREATION

Chronicles of Creation

Selected works bridging intelligence, threat analysis, and platform architecture.

RAG_Similarity_Engine.py
python query_rag.py --query "security threat"
[QUERY] "security threat"
[DATABASE] Searching vector index mappings...
[MATRIX] Similarity: Cosine(Q, D12) = 0.9412
[SUCCESS] Context retrieved: "Port scan detected on subnet..."
01.

RAG-Based Knowledge Retrieval System

A production-ready Q&A engine built for private, sensitive documents. Utilizes LangChain and FastAPI to index, retrieve, and generate contextually accurate answers from massive text files.

  • LangChain
  • FastAPI
  • Python
  • LLMs

Enterprise documents contain private, highly sensitive information that cannot be shared with public APIs, preventing the use of standard search tools.

Engineered a robust retrieval-augmented generation (RAG) system utilizing LangChain for semantic indexing and vector search, built on FastAPI.

Delivered a secure, self-hosted search engine that retrieves precise context and answers queries from massive unstructured file logs in milliseconds.

02.

Real-time Log Threat Detection API

An end-to-end AI security system that uses a hybrid modeling approach (combining Isolation Forest and XGBoost) for real-time behavioral anomaly detection in network and server logs.

  • Machine Learning
  • Isolation Forest
  • XGBoost
  • API Development

Static firewalls and basic log filters fail to catch sophisticated, shifting behavioral attack patterns in server network logs.

Built a dual-stage model combining Isolation Forest for unsupervised anomaly search and XGBoost for structured behavior classification.

Created an API capable of processing thousands of incoming server events per second with high classification accuracy and low warning latency.

Doubt How to map nested routes in Express?
Lost & Found Keys found near Block C library entrance.
Complaint Water dispenser outlet leakage in lab corridor.
03.

Campus Connect

A full-stack social support portal built with the MERN stack. Designed to streamline campus communications, help students post doubts, track events, report lost & found items, and submit complaints.

  • React.js
  • Node.js
  • Express
  • MongoDB
04.

Customer Churn Prediction

Developed predictive data models to analyze churn risk and customer behavior. Leveraged advanced data analytics and predictive modeling techniques to extract actionable retention strategies.

  • Data Analytics
  • Machine Learning
  • Predictive Modeling
Console Idle. Hover over index slots to trace calculations...
05.

Cyber Security - Threat Hash Detection

An efficient cybersecurity detection filter that implements custom Hash Tables for high-speed, low-latency checking of log entries against threat signatures.

  • Algorithms
  • Hash Tables
  • Security Analysis
Molecular Scan
Absorption (HIA) --
Toxicity (hERG) --
Clearance (TDC) --
BBB Penetration --
06.

ChemXplore

An ADMET prediction platform leveraging Chemprop-RDKit machine learning models trained on Therapeutics Data Commons (TDC) datasets. Evaluates large-scale chemical libraries for drug discovery.

  • ADMET-AI
  • Chemprop
  • RDKit
  • Python
  • HTML
GARDEN OF CAPABILITIES — GARDEN OF CAPABILITIES — GARDEN OF CAPABILITIES — GARDEN OF CAPABILITIES — GARDEN OF CAPABILITIES — GARDEN OF CAPABILITIES

A Garden of Capabilities

Like stones carefully selected and placed in a Zen garden, each tool serves a precise, mindful purpose in my development workflow.

Language Python
Language Java & C
AI / ML Deep Learning
AI / ML OpenCV & Vision
Frameworks FastAPI
Tools Docker & AWS
Language Python
Language Java & C
AI / ML Deep Learning
AI / ML OpenCV & Vision
Frameworks FastAPI
Tools Docker & AWS
Language JS & SQL
AI / ML CNNs & Detection
AI / ML TensorFlow
Frameworks React & Node
Databases MySQL & Mongo
Tools Git & Linux
Language JS & SQL
AI / ML CNNs & Detection
AI / ML TensorFlow
Frameworks React & Node
Databases MySQL & Mongo
Tools Git & Linux

Pathways of Growth

A log of professional experience, academic foundations, and certified capabilities.

Professional Experience

Jan 2026 – May 2026

Vinks Technologies

.NET Full Stack Developer Intern

Developed enterprise applications following SDLC principles and implemented RESTful APIs using ASP.NET and SQL Server. Collaborated with Agile teams to debug software issues, perform testing, and improve performance.

Feb 2026 – May 2026

Princeton Smart Engineers

Data Analyst Intern

Developed Python-based data preprocessing workflows and analytical dashboards to support healthcare research. Performed feature engineering and statistical analysis on large datasets for decision support.

Academic & Achievements

2022 - 2026

BLDEA College of Engineering and Technology

BE Computer Science (Artificial Intelligence & Machine Learning) CGPA: 8.0/10

Achievements

  • 3× Hackathon Winner

    Led developer teams to build high-performance software models under constrained timelines.

  • 3rd Prize — Engineering Project Exhibition

    Awarded for engineering a functional machine learning deployment at the final year project gallery.

  • Top 5% Candidate Selection

    Selected out of 100+ applicants for technical proficiency during the Princeton Smart Engineers assessment.

Certifications

  • AWS Certified Cloud Practitioner
  • IBM Data Analysis with Python
  • Cisco Cybersecurity Essentials
  • Deloitte Data Analytics Job Simulation
  • JPMorgan Chase Software Engineering Job Simulation
  • Tata Group Generative AI Job Simulation

Let's build in silence

Interested in AI development, ML model engineering, or data automation pipelines? Let's connect and build systems that last.