Sohail Mohammad
Production AI and Adjacent Thoughts
Forward deployed engineer at Together AI. Notes on production AI and adjacent thoughts.
Together AIInference optimization and post-training with enterprise customers
AmazonAgent platforms for 1.5M+ employees
JPMorgan ChaseRAG system from 0 to 10k+ users
Wendy'sGenAI drive-thru systems
Jack HenryGPU infrastructure
Start Here
Flagship
Production Inference Economics
A field guide, calculator, and essay series for measuring AI systems by loaded cost per accepted result.
Book
The Honest Field Guide to Production Inference
How to reason about provider choice, routing, retries, quality gates, serving, and migration economics.
Tool
LCPR Calculator
Run the loaded-cost math on your own workload instead of trusting token-price theater.
Selected Work
-
2026-06-08
Inference Flexibility Starts in the APIA response to Mario Souto's clay jar essay: energy orchestration only works if inference systems expose workload class, latency tolerance, and value per accepted result.
-
2026-05-11
The Idiot Index for AI DeploymentThe gap between what a model is supposed to do and what it actually does in production is enormous. Forward-deployed engineering exists to close it. Here's why i left Amazon...
-
2026-02-05
Ray in Production: What Dozens of GPUs and a Lot of 3am Pages Taught MeReal production failures from running Ray at scale: lost training runs, enterprise network disasters, and cascade outages the documentation never warns you about.
-
2026-02-03
Optimizing vLLM at Production Scale: Lessons from Conversational AI InfrastructureMemory fragmentation, throughput cliffs, and quantization accuracy issues that only show up in production—lessons from running vLLM at scale for conversational AI.
Latest Research
see all →
-
2026-03-17
Persona Circuits: Progress & Findings (3-17-2026)Current-state synthesis of the persona-circuits project: robust steering and partial concentration support, but weaker-than-expected distinctness, necessity, and sufficiency...
-
2026-03-17
Persona Circuits: Exploring GLP ApplicationBranch report on using Generative Latent Priors (GLP) for activation repair in persona steering: public-checkpoint transfer failed in this setting, matched checkpoints were...
Latest Writings
see all →
-
2026-06-08
Inference Flexibility Starts in the APIA response to Mario Souto's clay jar essay: energy orchestration only works if inference systems expose workload class, latency tolerance, and value per accepted result.
-
2026-05-13
Goodput or It Didn't HappenGPU utilization can be 78% while 30% of requests fail SLO constraints. The goodput frontier test replaces single-number benchmarks with decision-grade surfaces that measure...
-
2026-05-11
The Idiot Index for AI DeploymentThe gap between what a model is supposed to do and what it actually does in production is enormous. Forward-deployed engineering exists to close it. Here's why i left Amazon...