Remote SOC, six weeks in
Six weeks of fully-remote SOC operations. The functional measures are holding; the team-side challenges are the harder ones; EmilyAI's role in operational continuity has been larger than I would have predicted.
Opening 2020
First Monday back. The Citrix ADC clean-up dominates the inbox, the company turns ten in April, and the regulatory and threat landscape ahead has more sharp edges than I would prefer.
BSides Manchester: notes from the talk
The BSides Manchester talk on the Emily research went well. Forty-five minutes, three case studies, and the right kind of question afterwards. Notes for the file.
Opening 2019
First Monday of the year. The product roadmap is more concrete than at any previous point, the GDPR enforcement landscape is firming up, and the company is at twenty-six people across two offices.
EmilyAI pilot data, four months in
Four months of pilot deployment across three customers. The operational data is consistent with the internal deployment; the customer-side findings are more interesting than the model-quality findings.
EmilyAI live in customer SOC
EmilyAI commercial v1 went live on the first customer pilot deployment yesterday. The product is real. The internal name has stuck on the outside and the team's two years of work is, finally, in a customer's production environment as a product rather than as a service.
Opening 2018
First Monday back, and the year ahead is unusually well-defined. GDPR May 25, EmilyAI to production in Q2, the longer-form essay file finally moving into actual prose.
Emily, assist phase, six months in
Six months of the assist deployment. The numbers are interesting in directions I did not predict in January, and the analyst-team feedback is the more important data than the quantitative metrics.
Opening 2017
First Monday back. The shape of the year ahead — GDPR run-up enters its operational phase, Emily moves from shadow to assist, and the post-DNC threat landscape is the new normal.
First notes on machine learning for alert triage
First few weeks of an experiment to train a classifier on our analysts' triage decisions. The early results are interesting in ways I did not expect.