AI Memory Works · Author
Mrunmay Phanse — Author, AI Memory Works
Mrunmay Phanse writes AI Memory Works — practitioner guides on memory for AI agents, frameworks and architecture for developers building stateful agents.
Bio
About the author
Web developer and AI practitioner — building agent systems, memory architecture and developer tooling since 2014.
Mrunmay Phanse is a web developer and AI automation expert based in Pune, India. He builds custom React and Next.js applications, AI agent workflows and search-optimized knowledge bases. He founded CtrlDigit (digital marketing and web development) and Tabbly (AI-powered restaurant menus), and writes technical content on agent memory, Weaviate Engram and AI-native architecture.
AI Memory Works is written from a builder’s perspective: how memory layers integrate into production stacks, what LOCOMO and LongMemEval scores mean in practice, and when to choose Engram, Mem0, Zep or Letta.
Perspective
Expertise and editorial perspective
Practitioner voice, neutral comparisons, benchmark literacy.
- Agent memory architecture — memory layers, extraction pipelines, retrieval orchestration
- Framework evaluation — Engram, Mem0, Zep, Letta, LangMem with sourced benchmark numbers only
- Production patterns — user_id scoping, LOCOMO in CI, token cost optimization
- Neutral comparisons — no vendor pay-to-rank; gaps marked N/A when benchmarks aren’t published
Profiles
Publications and profiles
- mrunmay.dev — personal site
- GitHub
- DEV Community
Recent writing: Structuring Raw Interaction Data in AI Agents using Weaviate Engram (DEV Community)
Archive
Articles by Mrunmay Phanse
Flagship guides and hubs on AI Memory Works.
FAQ
Frequently asked questions
Who runs AI Memory Works?
Mrunmay Phanse writes and maintains AI Memory Works. See about page for editorial mission and comparison policy.
Is the author affiliated with Mem0, Zep, Letta or Weaviate?
No vendor affiliation. Tool comparisons use published data only. Engram is covered where it fits Weaviate-native stacks.
How do I contact the author?
Use the contact page for benchmark corrections, content feedback or general inquiries.
Speaking or consulting?
For speaking, consulting or collaboration, reach out via contact or mrunmay.dev.
How should I cite AI Memory Works?
Link to the specific page and cite the original benchmark papers (Chhikara et al., 2025; Rasmussen et al., 2025; Packer et al., 2023) for numbers. Author: Mrunmay Phanse, AI Memory Works.