welcome@whoami:~$ whoami
Shreyas Pandey
AIML Engineer
welcome@whoami:~$ cat skills.txt
welcome@whoami:~$ ls Experiences
welcome@whoami:~$ ls projects
welcome@whoami:~$ cat shreyas-skills.md
github.com/Shreyas-prog108/shreyas-skills
welcome@whoami:~$ ls blogs
welcome@whoami:~$ cat contact.txt
welcome@whoami:~$_
About Me
I'm an AIML Engineer focused on agentic systems, RAG pipelines, and production AI infrastructure. I work with LangChain, LangGraph, MCP servers, multi-agent frameworks, and full-stack tooling to ship scalable AI products.
I also maintain shreyas-skills — an open-source engineering framework (Karpathy-inspired) that gives Cursor, Claude Code, Codex CLI, Windsurf, and Copilot a shared standard for building agentic AI systems.
When I'm not building, you'll find me contributing to open source or writing on AI engineering.
Highlights
shreyas-skills
Open SourcePersonal engineering skill framework for agentic AI development — inspired by Karpathy's four LLM coding principles. One canonical standard for planning, building, and shipping agents across Cursor, Claude Code, Codex CLI, Windsurf, and GitHub Copilot. MIT licensed — fork and adapt.
REPPL
BuildingBefore the noise, see the signal. A dark, calm market read for founders who need to feel the shift before it feels obvious.
Lumicode
MCP ServerCustom MCP server for document extraction and LLM coding context — simplifies pulling structured context into agentic coding workflows.
Skills
Experience
Cellyn
AI Engineer Intern
Remote- Developed an enterprise-grade Excel add-in with agentic tool-calling infrastructure for real-time financial data modeling, enabling LLM-driven workflows to interact with structured financial datasets and analytical tools.
- Architected a private, encrypted data warehouse with access controls and secure data pipelines for sensitive financial information processing.
- Built an end-to-end enterprise Data Room platform from scratch in 20 days, featuring multi-tenant isolation, room-based access control, real-time notifications, audit logging, and Q&A capabilities.
- Engineered AI-powered features including context-aware memory management for chat generation, automated PPT generation, sandboxed code execution, web search integration, and multi-database connectivity.
- Managed full cloud infrastructure on Scaleway, owning CI/CD pipelines, deployment workflows, automated testing, and production releases across environments.
Fills AI
AI Engineer Intern
Remote- Engineered real time collaborative scientific editor (LangChain + Gemini + Tiptap + Y.js) with automated version control and cloud sync, processing 500+ biology manuscripts with 92% terminology consistency, reducing peer-review cycles by 40% for 50+ researchers.
- Built semantic RAG system over PubMed's 30M+ paper corpus (Sentence Transformers + vector embeddings) achieving 92% citation accuracy; added hands-free voice agent via Web Speech API, cutting drafting time by 40%.
- Integrated a live AI chatbot with real-time context extraction from dashboards and MCP server integrations, enabling agentic search across multiple platforms.
- Architected and deployed a custom MCP server for the company's internal agent, bridging separately hosted services with their platform for seamless agentic tool orchestration.
- Built a Biomni agent on Stanford's Biomni data lake, deployed on AWS EC2 with EFS-backed persistent storage for scalable biomedical data retrieval and analysis.
Stealth AI Startup
AI Engineer Intern
Remote- Engineered a real-time customer support voice pipeline using STT–TTS models, enabling low-latency conversational AI interactions for automated customer care services.
- Optimized end-to-end speech inference and streaming architecture, reducing system latency to ~36 ms through pipeline optimization, asynchronous processing, and efficient audio buffering.
- Designed and deployed the complete DevOps and inference infrastructure, orchestrating model services, real-time audio streaming, and scalable deployment on Scaleway cloud infrastructure.
21 Spheres
Python Developer Intern
Remote- Developed 5+ LangChain AI agents handling 1,000+ daily queries with 88% autonomous resolution rate across 5 client deployments.
- Architected multi-agent ReAct framework achieving 85% autonomous task completion and 60% reduction in client processing time and led 3-person engineering team end-to-end.
Featured Projects
Autonomous Data Analysis Agent
AI-driven analytical reasoning system with ReAct framework, natural language to code pipeline, and self-correcting multi-step analysis workflows.
Fin Alpha
Deep agent stock market intelligence system with live news triggers, sentiment analysis, and event-driven multi-agent workflows.
BlockSuite RAG Pipeline
Production-grade RAG with hybrid search, semantic deduplication, query rewriting, and PGVector retrieval across 1K+ documents.
Achievements & Leadership
YC Startup School 2026 — Top 1000 Builders in India
Shared engineering standard for agentic AI — works with Cursor, Claude Code, Codex CLI, Windsurf, and GitHub Copilot.
Open Source Contributor
Active contributor across 10+ repositories in Hacktoberfest, IWoC, and other OSS programs.