Talk to Parth about
How Parth thinks
Automate front‑desk
Reduces clinic front‑desk workload by handling appointment requests, rescheduling, and basic queries through an AI
For you: You can boost clinic efficiency with AI handling routine calls.
Memory‑efficient LLM serving
Identifies ways to shrink KV cache footprints, such as grouped‑query attention and paged attention, to enable more
For you: You can run more LLM queries on the same hardware by applying these cache tricks.
Low‑latency trading architecture
Designs distributed microservices in C++ and Go that execute trades within 100 ms, scaling to thousands of concurrent
For you: You can achieve fast trade execution by adopting similar service designs.
Ask this first
- How does the AI Receptionist handle insurance questions?
- What strategies do you use to keep KV cache memory low?
- Can you share challenges in building ultra‑low latency trading systems?
Posts
- Your LLM fits on the GPU and then 10 people start using it, and you run out of memory. The missing piece is the KV cache. Start from first principles: when a model generates the next token, that token needs to attend to earlier tokens. Recalculating the earlier tokens' keys and values every time wOct 13
- @Ishansharma7390 i remember people hating on ishan back in covid, and how he has come along so far is crazy tbhOct 12
- The foundary by Nikhil Kamath focuses on d2c brands, so, what is the YC alternative in India who is investing in technology/AI first startups?Oct 12
- @tokenaware i need to be back in blr frr now😔😔Oct 12
- It got to such large scale in India due to these factors- 1. People have fear of what will happen with them, they desperately need reassurance about the future 2. They want to get the best outcomes always- be it love, money, marriage, job, business. Hence, they believe in the products they sell. 3. Oct 11
- do you believe in starting an AI Services company right here in India and then target MSMEs in the US? I was working on the similar lines. i largely thought of 2 ways to approach this- 1. You make software solutions that these companies want with AI in fraction of TAT and cost perhaps. 2. You automOct 11
- Experimenting with performance marketing ad campaigns and strategies constantly, and that is the only way you make sure you do not lose in the game. Data analytics is the first thing you do after a week of marketing and understand what works and what doesn’t. You have to keep metrics that makes sensOct 11
- @bunmaskachaiii please don’tOct 11
- @lukepierceops bro i’ll sell your data bro, give it to me bro, 30% type shiiOct 11
- i’m sulking at 5 am listening to soul crushing musicOct 10
Parth's brain
Recently updated
I am developing an AI Receptionist voice agent for medical health centers and clinics, targeting the repetitive front‑desk workload and…
In early October 2026 I shared a series of X posts outlining the launch of an AI‑automation services startup aimed at US MSMEs, my funding…
I analyzed the memory pressure of the key‑value (KV) cache when serving large prompts on a GPU and identified three optimizations to reduce…
I implemented the TCP/IP stack in C/C++ with a CLI/Terminal interface that dumps ARP, MAC, routing tables, interface hops and graphs onto…