No. 01
POV
Why your forecast misses the spike
When one customer orders in bulk one month and nothing the next, accuracy at the tail beats accuracy on average. A field view from semiconductor planning.
Read →The physical world does not forgive vague software thinking. A bad decision shows up as a grounded aircraft or a stopped line, not a logged error. That hard constraint is exactly where Feasible does its best work.
The constrained systems where a better decision shows up on the floor or the schedule. Demand forecasting, network and fleet planning, downtime, and generative-AI copilots that read live operational data and act across the systems behind it.
Demand, pricing and assortment from production to shelf, with inventory and space optimization where margin is thin, and generative AI applied to assortment, content and customer engagement.
Practical AI in a high-stakes setting. Summarizing clinical and case notes, HR and member chatbots that actually answer, and forecasting where accuracy and trust both matter.
Predictive analytics on live competition. Win-probability models that re-score after every point, the performance keys players and coaches plan around, and scouting reports built from match footage in minutes. See the US Open case study.
When one customer orders in bulk one month and nothing the next, accuracy at the tail beats accuracy on average. A field view from semiconductor planning.
Read →The interesting work is not the chat box. It is the orchestration underneath, across the messy systems a real operation runs on.
Read →Moving a model from a promising AUC to something a clinician will actually act on, when the cost of a miss is measured in lives.
Read →How to run discovery jointly with a client team and separate the problems AI should touch from the ones it should leave alone.
Read →The most valuable thing an advisor can say is sometimes no. A short argument for killing the wrong project early.
Read →