Building AI-powered SaaS products, reliable backend systems, and intelligent workflows.

I'm a Product Engineer focused on building useful software products across backend engineering, full-stack development, and AI-powered applications. I enjoy turning complex requirements into maintainable systems—from API design and data modelling to AI workflow orchestration, integrations, and user-facing experiences.

• 🧠 AI engineering: LLM integrations, prompt engineering, agent workflows, RAG concepts, tool calling, and model evaluation.
• ⚙️ Backend engineering: Node.js, Express.js, Python, FastAPI, REST APIs, database design, caching, and background jobs.
• 🎨 Frontend engineering: React, Next.js, TypeScript, responsive interfaces, and reusable UI components.
• 🔄 Product integrations: Connecting external APIs and services for communication, payments, search, and automation.
• ☁️ Delivery and operations: Azure, CI/CD, process management, testing, and application monitoring.
• 🚀 Currently building: ResearchPilot, an independent AI research and decision-support agent.
• 📫 Email: praveen.pk200109@gmail.com
• 🌐 Portfolio: my-portfolio-reactjs-tailwind.netlify.app
• 💬 Let's connect: LinkedIn

JavaScript (ES6+) · TypeScript · Python · Java · HTML5 · CSS3

• React.js and Next.js
• TypeScript and modern JavaScript
• Tailwind CSS and responsive UI development
• Component-driven development and reusable UI patterns
• React Native exposure

• Node.js and Express.js
• Python and FastAPI
• REST API design, implementation, and integration
• MVC and service-oriented application structure
• Authentication, authorization, request validation, and error handling
• Asynchronous processing and background workers
• API integrations and service-to-service communication

• OpenAI API, Gemini API, and other LLM integrations
• LangChain and LangGraph
• Prompt engineering and structured output generation
• AI agents and multi-step workflow orchestration
• Retrieval-Augmented Generation (RAG) concepts and evidence retrieval
• Tool/function calling and external tool integration
• LLM evaluation and model benchmarking
• Local model experimentation with Ollama
• Guardrails, prompt-injection awareness, and input sanitization
• Token usage, latency, and cost tracking

• MongoDB and Mongoose
• MySQL and SQLite
• Redis caching
• BullMQ and queue-based background jobs
• Scheduled tasks with node-cron
• Data modelling, query design, and persistence

• Microsoft Azure
• Azure Virtual Machines and Azure Static Web Apps
• Docker and containerized development
• CI/CD pipelines and automated deployments
• PM2 process management
• Environment configuration and secrets management
• Application health checks, logs, and monitoring

• Twilio — communication and calling workflows
• SendGrid — transactional email delivery
• Stripe — payment integration
• Bright Data — web data collection
• Serper — web search integration
• GitHub API — repository and workflow integration
• OpenSearch — search and retrieval
• Third-party REST APIs and webhook-based integrations

• Jest
• Vitest
• React Testing Library
• Unit and integration testing
• API testing and debugging
• Regression testing and failure-path validation

• Git, GitHub, and Bitbucket
• VS Code
• Postman
• Debugging, code reviews, and collaborative development

ResearchPilot — AI Research & Decision Agent (Independent project · In development)

ResearchPilot is being designed to help users investigate complex questions, collect evidence from multiple sources, verify important claims, compare options, and produce structured reports with citations.

Researching a decision often means opening many pages, comparing inconsistent information, checking sources manually, and assembling findings into a report. ResearchPilot aims to make that process more systematic and auditable.

• 🧭 Research planning: Break a broad research goal into focused questions and subtasks.
• 🔎 Web research: Search public sources and retrieve relevant information.
• 🧾 Evidence records: Preserve claims, source URLs, excerpts, and retrieval timestamps.
• 🔍 Verification: Identify missing information, unsupported claims, and conflicting evidence.
• ⚖️ Comparison: Evaluate alternatives against explicit user-defined criteria.
• 📑 Decision-ready reports: Generate summaries, comparison tables, citations, risks, and uncertainties.
• 🛡️ Bounded workflows: Apply limits to retries, research rounds, execution time, and model usage.
• 📊 Evaluation: Test output quality, evidence coverage, latency, and cost.

The project uses three separate repositories: researchpilot-web, researchpilot-api, and researchpilot-agent.

Status note: ResearchPilot is under development. The capabilities above describe the intended product, not a claim that every feature is already implemented or deployed.

• Building maintainable backend services and clear API contracts.
• Designing data models and asynchronous job-processing workflows.
• Integrating LLMs into practical product features.
• Making AI outputs more structured, traceable, and testable.
• Integrating external services without tightly coupling the application to a provider.
• Improving reliability through validation, retries, testing, and observability.
• Learning through implementation, benchmarking, and iterative improvement.

Build thoughtfully. Learn continuously. Trust the process.