Packt Deep Engineering

Packt Deep Engineering

Home
Interviews
Newsletter Issues
Software Architecture
Engineering Leadership
Practical Deep-Dives
C++
Python
Sponsor Deep Engineering
About

Deep Engineering #71: Rohit Bhardwaj on Running Agentic AI Inference at Scale

On Per-Hop P99, GPU Queue Depth, and the 19 ROCS-Loop Patterns Behind Reliable Agentic Inference
READ THE LATEST
Packt Deep Engineering
Subscribe to Deep Engineering for weekly expert insights on software architecture, modern systems, and real-world engineering decisions.

Recent posts

View all
AI Inference at Scale
On the ROCS-Loop™, the 19 patterns that keep agentic inference reliable, observable, cost-aware and sustainable, and why your dashboards only show a…
Deep Engineering #70: SysML v2 and the Future of Systems Engineering with Dr. Bruce Douglass
On the textual notation that lets AI read and write systems models, and why better doesn’t mean simpler
 • Saqib Jan and Denim Pinto
Free ebook - How to differentiate between AI tools
Agents, assistants, co-pilots, orchestration…
Deep Engineering #69: Navigating Your Career as a Software Architect with Bruno Souza
On keeping your judgment as AI writes more of the code, and the career moves that build a reputation AI can’t fake
 • Saqib Jan and Srishty Goyal
Cognitive Surrender Is the Big Career Risk for Software Architects
On using AI for you instead of as you, and why architects must keep their judgment and their careers in their own hands
 • Saqib Jan
Deep Engineering #68: Fix the Workflow Before You Scale AI with Michelle Brenner
On why most companies see some AI return but few see a strong one, and the process map that closes the gap
 • Saqib Jan and Srishty Goyal
AI Adoption Pays Off Only When You Redesign the Workflow
On four leadership personas with competing AI goals, and the process map that turns tedious engineering work into agent tasks.
 • Srishty Goyal
30:53
AI Agents Need Governance Before Autonomy With Sibasis Padhi
Sibasis Padhi of Walmart Global Tech on why agents amplify retry storms, and why they should earn authority through a scored record before they act on…
 • Saqib Jan and Sibasis Padhi
Deep Engineering #67: Harness Engineering Patterns for AI Agents With Ali Arsanjani and Juan Pablo Bustos
As agents probe systems they were never meant to reach, the harness patterns that contain them, test them and keep scoring them in production
 • Saqib Jan and Ali Abidi
Governance and Harness Engineering Patterns for AI Agents
On instruction fidelity auditing, execution envelope isolation, trust decay scoring, and the maturity model that tells you which pattern to adopt next
 • Ali Abidi
Deep Engineering #66: Spec-Driven Development That Bounds AI Agents To Your Domain Model
Alessandro Colla and Alberto Acerbis on keeping AI coding agents from deciding what your domain never delegated
 • Saqib Jan
Spec-Driven Development Needs a Harness, Not Just a Spec
On domain carriers, guard agents, and catching the architectural drift an AI coding agent hides inside a green PR
 • Saqib Jan

Engineering Leadership

View all
Cognitive Surrender Is the Big Career Risk for Software Architects
On using AI for you instead of as you, and why architects must keep their judgment and their careers in their own hands
 • Saqib Jan
RelationalAI Measures Agent Readiness by the Friction Developers Absorb
On measuring agent readiness, shifting quality left of review, and keeping architectural intent where agents will find it
 • Saqib Jan
EY Bounds Agent Autonomy by Risk, Not by Capability
On why the agent boundary is a governance decision, and what has to exist before enterprise AI moves past the pilot
 • Saqib Jan
Engineering Teams Are Paying Back Their AI Speed Gains With Interest on Review
Agents are opening pull requests faster than many development teams can verify them, and engineering leaders are absorbing the difference in review…
 • Saqib Jan
Forward Deployed Engineer Jobs Are Out There. Hiring Is Harder Than The Postings Suggest
On hiring for AI’s messy last mile, structuring deployment teams, and turning customer lessons into reusable product improvements
 • Saqib Jan
Materials Science Gets Real Value From Quantum Before Anything Else
Small molecule chemistry follows close behind. Optimization, cryptography, and machine learning all need thousands of logical qubits, so those sit much…
 • Saqib Jan and Sebastian Hassinger
Eighty Years of Boolean Thinking Is the Real Barrier to Quantum Adoption
The hardware is not the hard part for enterprises. Recasting business problems in quantum terms means unlearning assumptions about problem solving that…
 • Saqib Jan and Sebastian Hassinger
Site Reliability Engineer (SRE) Interviews Measure a Depreciating Skill
Notable engineering leaders on motivation, ambiguity, tribal knowledge and the measurement problem behind every reliability hire that did not work out
 • Saqib Jan

Interviews

View all
30:53
AI Agents Need Governance Before Autonomy With Sibasis Padhi
Sibasis Padhi of Walmart Global Tech on why agents amplify retry storms, and why they should earn authority through a scored record before they act on…
 • Saqib Jan and Sibasis Padhi
37:35
How Cisco's Outshift Runs Engineering On AI Tiny Teams And A Memory Engine
Vijoy Pandey shrank his teams to five people and explains what that costs, and the memory engine and decision leads Outshift built to hold it together
 • Saqib Jan
Multi-Agent Architectures in Production with Chi Wang and Xiao Ma
When one agent stops being enough, how to tell over-decomposition from under-decomposition, why observability moved from operations into design, and…
 • Saqib Jan
Quantum Computing Beyond the Hype with Sebastian Hassinger
How to tell genuine quantum progress from hype, why qubit count misleads, where the technology delivers value first, and how a classical developer…
 • Saqib Jan and Sebastian Hassinger
Database Choice, Lock-In, and the AI Rush with Peter Zaitsev
On why nobody owning Postgres became its advantage, the labels that sell you proprietary software, and what self-hosting costs once AI speeds up…
 • Saqib Jan
Tokens, Judgment, and AI Productivity with Austin Parker
On why a token is not a unit of intelligence, the illegible work token counts miss, pointing AI budgets at developer experience and observability, and…
 • Saqib Jan
Recommendations
Agentic Engineering
Agentic Engineering
Packt
The latest from
Saqib Jan
Deep Engineering #71: Rohit Bhardwaj on Running Agentic AI Inference at Scale
User's avatar
Steven Lott
Deep Engineering #38: Steven Lott on Practical Object-Oriented Design in Python
User's avatar
Scott Millett
Deep Engineering #35: Scott Millett on closing the “story gap”
User's avatar

Subscribe

Deep Engineering is a weekly newsletter for developers and software architects featuring expert-led insights, deep dives into modern systems, and clear thinking on real-world software design.
AboutArchiveRecommendationsSitemap

Packt Deep Engineering

© 2026 Packt · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for writing, podcasting, and media founders.