Blog17 min read
Tool-Calling Reliability: 90% Per Call = 35% Over 10 Steps
LangGraph, CrewAI, AutoGen and LlamaIndex compared on tool selection, parameter accuracy, error recovery and retries, and why step count compounds failure.
Read article
Blog
Notes from the build.
Field notes on shipping AI-native systems with human engineering judgement: architecture, modernisation, automation, and the tradeoffs in between.
Blog18 min read
LangGraph State Management: What the Documentation Doesn’t Cover Until You’re Already Committed
State is a contract: use durable checkpointers, reducers, pruning, and schema versioning to make LangGraph agents production-safe.
Read article
Blog13 min read
Agent Benchmarks Lie: 95% Per Step Fails 40% of the Time
Benchmarks test one perfect run; production needs consistency. Pass-consistency, chaos testing, trajectory validation and p95/p99 latency explained.
Read article
Blog19 min read
How to Choose an Agent Stack: An Evaluation Framework and a 2-Week Spike
Criteria and a two-week test plan for picking an AI agent framework: reliability, state persistence, observability, coordination, and cost.
Read article
Blog19 min read
Multilingual Voice Agents: Keeping Latency Under 1,000ms
Language detection, STT model choice and a three-tier fallback for multilingual voice agents. Whisper Large-V3 for accuracy, Vosk offline, under 1,000ms.
Read article
Blog32 min read
Voice Agent Production Readiness Checklist: 11 Things to Stress Test Before Enterprise Deployment
11 critical stress tests to ensure enterprise voice agents handle noise, interruptions, networks, PII, load, and failover.
Read article
Blog14 min read
Why VAD End-of-Speech Detection Is the Hardest Problem in Production Voice Agents
Why VAD fails in real-world voice agents and how hybrid acoustic+semantic methods reduce cutoffs, false triggers, and latency.
Read article
Blog11 min read
Voice Agent Latency: Where the 2-3 Second Delay Actually Lives in the Pipeline and How to Reduce It
Break down where 2-3s voice-agent lag comes from and how streaming VAD/STT/LLM/TTS plus WebRTC cuts perceived latency.
Read article
Blog17 min read
Document AI Fails at Extraction, Not at the Model
When Document AI misreads $38,000,000 as $88,000,000 the model is fine and the extraction layer is broken. Where OCR, reading order and tables go wrong.
Read article
Blog18 min read
LangGraph vs CrewAI vs AutoGen: State, Cost, and Reliability Compared
How LangGraph, CrewAI, and AutoGen compare on the things that decide a production deployment: state persistence, cost per task, failure behaviour, and scalability.
Read article
Blog19 min read
Ship Fast Without the Fallout: A Framework for Tech Debt at Every Stage
A stage-by-stage approach to technical debt across validation, growth, and scale: what to let slide, what to refactor, and what blocks enterprise readiness.
Read article
Blog14 min read
What Every Experienced CTO Wishes They’d Done Differently About Technical Debt
Technical debt is a deliberate tool - manage, measure, and repay it with traffic-light roadmaps, AI detection, and business-aligned timing.
Read article