WorkExperienceWritingRésuméMessage me on LinkedIn
Portrait of Satyam ChatrolaSatyam Chatrola

I make LLMs fast,
reliable, and measured.

AI Systems Engineer at Mursion, specializing in LLM inference, reliability, and evals. I serve models with low tail latency, keep them up under load, and measure that their answers hold up.

Inference
71%

p99 latency cut in production

Reliability
355 MB

Deploy image, down from 14 GB

Retrieval
5M+

SKUs served by hybrid retrieval

Retrieval
+34%

Click through rate lift

Three specialties

Fast to answer,
hard to break, measured to be right

Inference

Batching, KV cache reuse, quantization, and semantic caching, so the first token arrives sooner and every token after it costs less.

Reliability

Async load shedding, backpressure, and circuit breakers under heavy traffic, incident root cause work, and drift monitoring that catches a decaying model early.

Evals

I build the evals that decide which model ships: candidate models are tested against product criteria and performance, side by side, before one is chosen.

A model is only as good as its slowest response and its worst answer. I build for both.

Experience

Production work, measured

AI Systems Engineer

Mursion

My work sits where async systems, concurrency, and ML serving meet. I am going deeper into LLM serving: throughput, KV cache efficiency, and tail latency.

Python asynciovLLMDockerArgoCDFAISS
Current

Building evals for model decisions: candidate models are tested against product criteria and performance before one is chosen.

71%

Cut p99 latency by moving blocking calls off the asyncio event loop, while keeping session order intact.

Async

Kept throughput steady under heavy concurrent load with an async logging queue: watermark based load shedding and backpressure to reduce GIL contention.

<7 min

Rebuilt multi stage Docker builds. The image went from 14 GB to 355 MB, and PR to production went from 30 minutes to under 7.

Root cause

Found a silent misrouting between builds that caused a production traffic skew during a Jenkins to ArgoCD migration. Engineering leadership recognized the fix.

$300K+

Built hybrid retrieval (two tower network, FAISS, and BM25) over 5M+ SKUs. Conversion rose 7.2% and click through rose 34%, driving $300K+ ARR.

Hiring for inference, reliability, or evals work?

Tell me about the team and the problem you want to solve.

Message me on LinkedIn