Tanishq
Kumar
AI Deployed Engineer / Software Architect / Founding Systems Developer
I built the agent harness behind from scratch. Agent frameworks are excellent for prototyping; in production, when I own the requirements and understand the intent, a purpose-built system comes out leaner, faster and better optimized.
What the work has moved
- 15 min → <100 ms
- pipeline compilation after replacing LLM assembly with an AST compiler, roughly 3,000× faster
- 2 h → <5 min
- Kubernetes cluster provisioning through modular Terraform
- 99.9%
- production uptime behind Cloudflare edge routing
- 84%
- prompt-cache hit rate on OpenAI from structured prompt prefixes
- SOC 2 Type II
- compliance held with an automated audit and scanning bench
Now
Founding engineer at , building the platform since June 2025. Day to day: autonomous agent architectures, high-throughput backends in Python and Rust, and cloud-native infrastructure.
Founding backend and AI systems engineer with 2.2+ years of production experience architecting enterprise GenAI platforms and agentic AI systems, distributed microservices, and automated cloud infrastructure. The harness runs on MCP servers and unified agent tools I set up, in the app or over MCP, and on a custom AST compilation engine I wrote. I also translate business-level data semantics and data contracts into technical models, bridging frontier LLM research with high-reliability enterprise backend engineering.