Talk to Harkirat about
- AI compliance for crypto
- Agent memory architecture
- Decentralized trading platform
How Harkirat thinks
Compliance as built‑in
Sees compliance as a mandatory part of AI product workflows, enforced before messages are sent.
For you: Consider embedding compliance checks early in your AI pipelines.
Memory is a system design
Views agent memory as four distinct layers that must be managed to keep long sessions efficient.
For you: Structure your agent’s memory into working, semantic, episodic, and procedural parts.
Productionizing AI agents
Believes moving AI agents from development to production is the hardest step, prompting the creation of neatlogs.
For you: Look for platforms that streamline the deployment of AI agents.
Ask this first
- How do you integrate compliance checks into AI-generated content?
- Can you explain the four‑layer memory model for long‑running agents?
- What challenges did you face launching neatlogs for agent production?
Harkirat's brain
Recently updated
On 2026‑09‑24 I tested ZeroDrift’s compliance sandbox with a crypto‑marketing claim while building my decentralized trading platform.
In September 2026 I posted several X (Twitter) updates covering a developer talk recommendation, a Solana open‑house announcement…
Launched @neatlogs to ease moving AI agents from development to production (see https://t.co/IwhlPk2kR1).
I am a software professional based in Bengaluru, Karnataka, India, with a B.Tech from IIT Roorkee (2014‑2018). My career spans finance…