Python Developer building AI data tools, automation agents, and Linux/DevOps diagnostics.

Projects | LinkedIn | Email

I build practical software projects around AI agents, data analysis, Linux operations, automation, and infrastructure tooling. My current focus is on portfolio-grade systems that accept messy inputs, run real analysis, and produce useful reports or actions.

• Working on: AI data apps, autonomous agents, Ubuntu diagnostics, and DevOps automation
• Comfortable with: Python, React, pandas, scikit-learn, Streamlit, Linux, Bash, Docker, GitHub Actions, and APIs
• Interested in: Linux systems, SRE tooling, AI engineering, backend systems, and cloud infrastructure
• Goal: build reliable tools that turn raw data, logs, and user questions into clear decisions

An AI-powered credit risk assessment and regulatory compliance platform that retrieves regulatory rules from a FAISS vector store, performs structured evaluations using the Claude API, and validates outputs using an LLM-as-a-Judge framework.

What it does

• Automated Financial Ratios: Computes Current Ratio, Debt-to-Equity, Interest Coverage, and the Altman Z-Score from balance sheets.
• RAG Regulatory Retrieval: Indexes Basel III, IFRS 9, and SEC regulations into a FAISS vector store using LangChain.
• Prompt Engineering Workshop: Compares Zero-Shot, Few-Shot, and Chain-of-Thought prompts side-by-side.
• LLM-as-a-Judge Audits: Evaluates assessments for accuracy, completeness, and regulatory alignment, logging results to track accuracy gains.
• Interactive UI & RBAC: Provides a glassmorphic React (Vite) dashboard with user logins, role access (ANALYST / ADMIN), and automated pytest coverage.

Tech: Python, FastAPI, React, Vite, SQLAlchemy, LangChain, FAISS, Anthropic (Claude API), pytest, CSS

Repo: financial-risk-analysis-ai

A production-grade microservices banking backend system that handles customer account management, transaction processing (debits, credits, transfers), and automatic transaction fraud risk evaluations.

What it does

• State Machine Transactions: Implements transition stages (PENDING -> PROCESSING -> COMPLETED/FAILED) with immutable PostgreSQL event log records.
• Idempotency & Concurrency: Restores cached responses using @Idempotent aspects on HTTP Idempotency-Key headers, and uses @Version optimistic locking with lock-ordered Account IDs to prevent balance deadlocks.
• 8-Signal Fraud Rule Engine: Evaluates transaction velocity, off-hours, location anomalies, young accounts, new payees, and device fingerprints to generate risk scores (0-100).
• Compliance Hold Queue: Routes high-risk transfers ($\ge 70$) into a review hold queue, running automated natural language risk explanations.
• Automated CI/CD & Deploy: Configured with Docker Compose orchestrators, Dockerfiles, and GitHub Actions Maven unit/mock test validation.

Tech: Java, Spring Boot, Spring Security (JWT), Hibernate, Flyway, PostgreSQL, H2, JUnit 5, Mockito, AOP, Docker, GitHub Actions

Repo: banking-system

An interactive terminal and Streamlit-based incident response agent that merges safe local systems tool execution with LLM reasoning.

What it does

• Safe structured systems diagnostics (disk metrics, systemd unit loops, network listeners, process hogs, and journal logs)
• Supports offline SRE simulation mode as well as live Gemini and OpenAI API function-calling loops
• Generates safe-to-execute Bash mitigation scripts with comprehensive explanations
• Fully verified with active Python testing suites and structured GitHub Actions CI automation

Tech: Python, systemd, journald, Streamlit, Rich CLI, pytest, Google Gemini API, OpenAI API

Repo: ai-sre-agent

An AI-powered Streamlit app that turns a CSV upload and a business question into dataset profiling, cleaning, baseline ML models, automatic charts, and a Markdown analysis report.

What it does

• Uploads and profiles CSV datasets
• Cleans common tabular data issues
• Infers target columns and classification/regression tasks
• Trains baseline scikit-learn models
• Generates Plotly charts automatically
• Writes a business-style analysis report
• Includes tests, sample data, and a 22-commit project history

Tech: Python, pandas, scikit-learn, Plotly, Streamlit, pytest

Repo: ai-autonomous-data-scientist

A Linux/SRE-style diagnostics toolkit that collects system evidence, detects common failure patterns, recommends fixes, and generates incident reports.

What it does

• Diagnoses failed systemd services and journalctl logs
• Detects service issues like port conflicts, restart loops, config errors, missing files, and permission failures
• Analyzes disk-full incidents, deleted-but-open files, and Docker disk usage
• Parses SSH/auth logs for brute-force attempts, successful logins, and sudo activity
• Collects system health from /proc
• Includes CLI, Streamlit dashboard, sample logs, tests, and report generation

Tech: Python, Linux /proc, systemd, journald, Docker, Bash, Streamlit, unittest

Repo: ubuntuops-agent

An AI/ML-style analyst agent that turns a raw CSV/XLSX dataset and a natural-language question into a structured insight report.

What it does

• Profiles uploaded datasets
• Cleans columns, dates, missing values, and duplicates
• Plans analysis from a user question
• Generates pandas-based insights
• Creates visual charts
• Exports an executive Markdown report
• Includes CLI, Streamlit UI, sample data, and tests

Tech: Python, pandas, NumPy, matplotlib, Streamlit, unittest

Repo: insightforge-agent

Languages

AI / Data

DevOps / Systems

• AI agents that use tools, files, APIs, and code execution
• Data apps that convert raw datasets into clear reports
• Automation scripts for repetitive engineering workflows
• Linux and DevOps projects with real debugging, logs, and CI/CD
• Backend tools with clean APIs and practical documentation

AI data apps + Python automation + Linux diagnostics + DevOps tooling

I am actively building projects that are easy to run, easy to verify, and useful enough to explain in interviews.