Hi, I'm Sandipan Paul — Full Stack Engineer | Cloud & Platform

Building reliable software and scalable systems.

Architecting resilient systems from chaos. Specialized in Full Stack Engineering, Cloud & Platform Infrastructure, and scalable web applications.

ACTIVE PROCESSES
> Backend-Services.init()
> Agent-Orchestrator.boot()
> Kubernetes-Cluster.scale(nodes=5)
> Observability-Daemon.trace(latency=42ms)
> Cloud-Infrastructure.sync(status=healthy)
> System Status: RUNNING

~/PROJECTS

EvalFlow

[Evaluation] [LLM Systems] [MCP Server] [SDK] [Tracing]

Open-source LLM evaluation toolkit and backend-first eval platform with an MCP server, SDK, prompt versioning, benchmark runs, and LLM-as-judge scoring.

CodeCortex AI

[Python] [LangGraph] [Groq] [GitHub]

Autonomous AI code review agent for pull requests, focused on bug detection, regression checks, and security-oriented reasoning.

FrameForge

[LangGraph] [VideoDB] [Groq] [Agents] [Python]

Autonomous AI video production workflow using LangGraph, VideoDB, and Groq for prompt-driven scene planning, retrieval, orchestration, and automated production pipelines.

Observa

[MLOps] [DVC] [MLflow] [Grafana] [NLP]

End-to-end NLP and MLOps monitoring and observability pipeline with experiment tracking, metrics dashboards, artifact versioning, and telemetry for model visibility.

~/PAPERS

Attention Is All You Need

Vaswani et al. | NeurIPS 2017 | Transformer implementation study

Rebuilt the core transformer architecture to study attention mechanics, encoder-decoder flow, and training behavior from scratch.

LeNet vs AlexNet: CNN Architecture Comparison

LeCun et al. + Krizhevsky et al. | Comparative CNN study

Implemented the original convolutional pipeline to understand early vision architectures, parameter sharing, and classifier construction.

~/EXP

Software Engineer (Cloud & Platform) | Kudrat Foundation

Sep 2026 - Present

AI/ML Engineer | CRID India

May 2026 - Sep 2026
  • Built an unattended Azure pipeline combining bilingual (Hindi + English) ASR and LLM summarisation, turning Fireflies recordings into structured Minutes of Meeting in SharePoint within ~5 minutes of a meeting ending.
  • Built a contact-research tool that turns an organisation name into a verified list of its board and leadership in about a minute, using multi-provider search, LLM failover, and evidence-based matching.
  • Delivered, with the team, AI Auditor and Legal Advisor demo platforms for the Bureau of Energy Efficiency (BEE), Ministry of Power, under the ADEETIE scheme.
  • Contributed to the AI Auditor workflow that generated DPR drafts in minutes, reducing manual audit report effort.
  • Supported Legal Advisor case-tracking workflows by structuring legal information and document review flows.
  • Fixed and deployed CRID India's website with SEO, analytics, Resend email forms, and a Google Sheets-based no-code CMS, driving 1,300+ page views and 1.2K%+ growth.

AI & ML Intern | Strydden Technologies

June 2025 - May 2026
  • Engineered Jyotirva, an AI-assisted astrology backend using Python, FastAPI, and Swiss Ephemeris with 15+ REST APIs across natal charts, Panchang, dasha, compatibility, divisional charts, yogas, and transits; improved accuracy across 1,100+ validated outputs with validation workflows for 17 endpoints and 59 passing backend tests.
  • Engineered a production-ready FastAPI clinical triage service using RAG + Pinecone for high-relevance symptom retrieval.
  • Delivered persistent multi-turn patient conversations with MongoDB session management and WATI WhatsApp webhook orchestration.
  • Architected an ontology-driven symptom taxonomy and tuned semantic search to improve retrieval relevance.
  • Implemented tamper-evident medical report signing with SHA-256 + ECDSA and built an automated security test suite.

Robotics Intern | Just Robotics

July 2024 - July 2025
  • Engineered RC control interfaces over websockets and serial transport with lower command latency.
  • Built a lightweight Flask backend for real-time control message handling.
  • Focused on responsive tooling for hardware-in-the-loop interaction.

~/BLOGS

~/CONTACT

Reach out here if you want to collaborate, talk backend systems, cloud infrastructure, or build something useful together.