Team
Co-founder & Engineering Lead
Irfan
Architecting sub-millisecond semantic vector RAM, distributed wire-level proxy engines, and closed-loop continuous model distillation queues.
- 2
- Research works
- 99.4%
- Parity win-rate
- 1.8 ms
- P50 Latency
- 78%
- Cost savings
- October 2026
- Cohort
About
Architecting sub-millisecond semantic vector RAM, distributed wire-level proxy engines, and closed-loop continuous model distillation queues.
Obsessed with P99 sub-2ms cache lookups, zero-waste wire runtimes, and regex grammar compliance.
Writes about
Research & Notes 2
- Preventing Unbounded LLM Consumption Mitigate OWASP LLM10 Unbounded Consumption risks with layered controls across edge, orchestration, quotas, and real-time proxy budgets.
- Constrained Decoding and Structured Model Output Understand schema-constrained output, JSON mode, reliability boundaries, retry economics, and validation stages for production LLMs.
- Sub-2ms Vector RAM Telemetry and Regex Compliance Empirical measurement of 99.8% JSON schema compliance with zero client-side degradation.