Anup BhowmikAI Enthusiast & Researcher

Assist in teaching a course focusing on data structures and algorithms, using C++ as the primary language.

Offer one-on-one help during office hours and review student code to improve logic and efficiency.

Single-handedly developed AI Pipelines for a cloud-based ERP system.

Capabilities:

• Extract features from Natural Language Query (with live voice transcripts) and parse precisely. Generate Structured Output for seamless integration with frontend using Pydantic Schema.
• Generate accurate SQL queries from Natural Language and show visualization using plotly.
• Retrieval-Augmented Generation (RAG) using LangChain, enabling real-time data retrieval and enhancing response accuracy in a data-heavy ERP system.
• Self-hosted local Large Language Models (LLMs) using Ollama, reducing dependency on external APIs and improving data privacy.
• AI-driven business insight generation system, leveraging machine learning and Business Intelligence(BI) tools to provide real-time analytics, reports and decision support.
• Scraping and data extraction from webpages and curate meaningful information using LLM
• LLM analytics and tracing with a self-hosted Langfuse instance, improving observability, performance monitoring and cost analysis.

Deployed the backend and database of the cloud ERP on ACS (Azure Cloud Service) using AKS (Azure Kubernetes Service), ACR(Azure Container Registry), and Cloud storage.

Incorporated CI/CD with GitHub actions, accelerating integration and deployment across multiple microservices.

Led the development of frontend (React) application and Landing page (Next.js), focusing on performance and user experience. Deployed the Webapp in Azure Static Web Apps.

Mentored a team of 4 trainee software engineers, guiding them in software development best practices and project workflows.

Conducted multiple Training Sessions on Java as part of the EDGE Project, delivering technical instructions to Trainee Developers.

Implemented CI/CD pipelines using Watchtower, Docker, and GitHub Actions, automating image builds, pushing to registry, and cloud deployments, reducing deployment time by 90% and increasing system reliability. I've written a Hands-on Guide on how to set up this automation, click to read the Medium Article.

Conducted load testing and performance analysis using k6. Identified bottleneck APIs and enhanced system scalability by 80% by suggesting new cloud infrastructure design.

Transformed a single-VM setup into a scalable distributed system with Nginx, Redis and Network File Share, achieving 40x performance improvement. Scaled to 2000+ users with 850ms latency and 99.9% uptime. Enhanced security by setting up an internal service communication network. Here is the detailed system design Article.

Configured Nginx as a reverse proxy and automated SSL certificate management with Let's Encrypt and certbot, improving security and reducing manual intervention.

Managed DNS configurations in Cloudflare, optimizing domain security, traffic routing, and reducing latency.

Developed a web-based early warning system to alert erosion-prone coastal households in Bangladesh, potentially safeguarding 1000+ residents and livestock from riverbank erosion hazards.

Integrated Google Maps API to visualize real-time riverbank erosion data, providing users with up-to-date geographical hazard alerts.

Published a research paper on the system’s effectiveness and implementation. Read the Paper

Deployed and maintained the live system, accessible at ews-re.com.