MechAI · Adam Krysztopa
Where the physicalworld meetsartificial intelligence.
I help industrial teams turn sensor and process data into models their operators actually trust and run — predictive maintenance, digital twins, forecasting, and RAG / agentic systems. PhD physicist and ex–GE Aerospace engineer, now an AI/ML team lead bridging physical systems and production-grade AI.
- Heat Transfer
- GE Aerospace
- Predictive Maint.
- RUL & twins
- Forecasting
- time series
- Production AI
- RAG & agents
An engineer who learned to teach machines.
I'm a PhD physicist (physics of semiconductors) with an engineering background from GE Aerospace, where I led heat-transfer and secondary-flow design on the GE Catalyst, ATP, and CF6 programs — environments where a wrong assumption melts hardware, not a unit test. That respect for physical limits never left me.
Today I bridge complex physical systems and production-grade AI: predictive maintenance, digital twins, time-series forecasting, and production-ready RAG and agentic systems — turning domain expertise into scalable, explainable solutions for industrial environments at STX Next.
MechAI is my own practice for engineering-aware analytics — Oil & Gas reliability, Remaining Useful Life, digital twins, and forecastability triage. I'm the author of dependence-forecastability, an open-source toolkit for telling whether a time series is even worth forecasting before the expensive modeling begins.
View full CV→Operating principles
- Physics firstConstraints are features, not obstacles.
- Ship the loopA model in production beats a notebook PB.
- Explainable by designIf an engineer can't trust it, it won't run.
- Measure the real costLatency, energy, and failure modes count.
From a thermal boundary to a working model — one circuit.
- IN01
Heat Transfer
Thermal systems and secondary-flow design at GE Aerospace — Catalyst, ATP, CF6. Where physical limits are non-negotiable.
- Thermal
- Secondary Flow
- GE
- x²02
Forecasting
Time-series and probabilistic forecasting — plus forecastability triage to know what's learnable before modeling begins.
- Time Series
- Probabilistic
- Triage
- ∇03
Predictive Maintenance
RUL, reliability, and digital twins for industrial assets — interpretable models operators actually trust.
- RUL
- Digital Twins
- Reliability
- AI04
Production AI
Elastic RAG and agentic systems — explainable, observable, and built to ship in regulated environments.
- RAG
- Agents
- LLMOps
Tools I've built for forecasting, reliability, and applied AI.
Writing from the seam between physics and machine learning.
- Oct 27, 2026Your Agent Can't Grade Its Own Homework: What Reflexion Actually ShowedAn agent improves only as far as a signal it does not produce lets it. Reflexion's gains rested mainly on the pass/fail check, not the reflection, and plain retry beat it for less money.#Agents#AIEngineering#Research10 min→
- Oct 13, 2026MechAI Is on YouTubeI'm starting a YouTube channel. Tech in general, but mostly aviation: how aircraft systems work, and where AI actually helps keep them flying.#Aviation#YouTube#Announcement1 min→
- Sep 18, 2026The Fix Is the PatternA pattern name buys you nothing. The step budget, the observation check, the reject path, and the human interrupt are what actually stop an agent from looping, leaking, or lying. 46 LangGraph traces, with token counts from the traces.#Agents#AIEngineering#Architecture#LangGraph16 min→
- Aug 20, 2026Show Me Your Graph. Tell Me You're a Rookie.A graph is not an architecture, it is a control flow representation. When an LLM agent framework becomes the default shape, the diagram answers the design question for you.#Architecture#AIEngineering#Agents11 min→
- Jul 31, 2026Choose Evidence, Not a ShapeHow to design a testing strategy from risks, evidence, and cost — not pyramids, trophies, or prescribed ratios.#Testing#Quality#Interactive17 min→
- Jun 14, 2026arch-crew: a plugin for deciding (and auditing) architectureThree Claude Code skills that help you pick — or assess — software architecture, design patterns, and agentic systems, recommending the least architecture that meets the requirement. They grew out of three interactive decision wizards you can click through here.#Architecture#Claude#Interactive4 min→
- May 28, 2026Is It Even Forecastable? Triage Before You ModelMost forecasting projects start with model selection. They should start with a cheaper question — does this series contain exploitable structure at the horizon you care about?#Forecastability#TimeSeries#OpenSource2 min→
- May 12, 2026A Digital Twin Is Not a Fancy DashboardIn manufacturing, most 'digital twins' are real-time monitoring dressed up with 3D visuals. A real twin answers a different question: what happens before you change anything?#DigitalTwins#PredictiveMaintenance#Manufacturing2 min→
- Apr 15, 2026Remaining Useful Life: The Model Is the Easy PartRUL estimation for industrial assets looks like a regression problem. The hard parts are the label, the cost of being wrong, and whether an operator will trust the number.#PredictiveMaintenance#RUL#Reliability2 min→
- Mar 2, 2026From Heat Sinks to Backprop: An Engineer's Path into AIThe transition from mechanical engineering to AI isn't a pivot away from physics — it's physics gaining a new instrument.#Career#Engineering#Reflection1 min→
Have a problem where physics and AI collide?
Predictive maintenance, digital twins, time-series forecasting, or production RAG and agentic systems — or just figuring out whether ML is even the right tool — let's talk.