Actian hires pg-jev creator Mahmoud Zachi for its enterprise AI push
The German indie developer says a September 17th post about plain-language Postgres search changed his life and led to the move.
By Ryan Merket · Published
Primary source: X
Why it matters
Actian's hire shows how quickly an open-source AI prototype can become a recruiting signal when it maps a new model onto software enterprises already use.
Mahmoud Zachi (@iam_zachi) has joined Actian to work on bringing AI into enterprise software, four days after publishing a PostgreSQL extension that lets users filter database rows with plain-English conditions.
https://twitter.com/iam_zachi/status/2100679300756435135
Zachi announced the move in a three-post thread on September 21st, calling the preceding weekend "crazy" and saying the original product post had changed his life. He did not describe a formal project, offering instead a concise mandate: bring cutting-edge AI into enterprise software quickly.
The timing makes the product behind the hire central to the story. On September 17th, Zachi posted a demonstration of pg-jev, an open-source extension that inserts model-based judgments directly into PostgreSQL queries. His examples turned descriptions such as "could work from home" or "name sounds European" into conditions inside an ordinary SQL WHERE clause.
An archive of the post recorded about 2,800 likes. By September 21st, the pg-jev repository had collected 267 GitHub stars and 14 forks. Those numbers establish early developer interest rather than production adoption, but the project gave Actian a working artifact to assess instead of a resume claim.
Zachi arrived at the demo through an unconventional route. On his personal site, he identifies himself as a software engineer and security researcher based in Gladbeck, Germany. He worked as a tax officer for North Rhine-Westphalia's finance ministry from 2018 to 2022, moved into enterprise Java engineering at KISTERS, and later ran IT for K.A.Y. Security Service. He became an independent builder in December 2025.
His recent work spans an AI-assisted App Store screenshot editor, a native Mac dictation tool, server-side Shopify analytics and a private journaling app for parents of autistic children. He has also published mobile security research, giving him experience with the failure modes that appear when software sends sensitive data into systems outside an operator's control.
From a two-day build cycle to enterprise software
Pg-jev appeared two days after TypeSafe AI introduced Jev on September 15th. TypeSafe founder Diogo Almeida described Jev as the first release in a class of "System One" models designed to return fast, structured decisions for software rather than generated prose.
Zachi wrapped that model in familiar database syntax. Pg-jev exposes functions for filtering, classification, ranking and probability scoring. A query can ask whether a support ticket sounds angry, decide which department should handle it, or order products by a subjective concept such as luxury.
The implementation makes deliberate trade-offs. Pg-jev performs a full scan and sends the relevant row contents to TypeSafe's API. Zachi's documentation warns against using it on data that cannot be shared with a third party. It also requires PostgreSQL's untrusted PL/Python extension and superuser access, excluding managed services such as Amazon RDS, Neon and Supabase under their standard configurations.
Zachi reports that pg-jev processed a 2,000-row test in about 3.5 seconds at an estimated cost of $0.012, while a repeated query completed in roughly 50 milliseconds using its session cache. Those measurements come from the project's author and depend on network conditions, row sizes, model pricing and query design. Pg-jev's significance lies in the interface: it turns model calls into composable database operations that existing SQL users can inspect and combine with conventional predicates.
That approach lines up with Actian's current product direction. Actian, the data and AI division of HCLSoftware, sells database, analytics, governance and observability products for large organizations. Its 2026 releases include a vector database, conversational analytics, data observability agents and an agentic data steward. Actian also absorbed AI analyst startup Wobby and embedded analytics vendor Jaspersoft into its portfolio.
Actian's challenge is converting model capabilities into features that can survive enterprise requirements around access control, auditability, deployment and data residency. Zachi's prototype does not solve those requirements on its own. It provides a compact demonstration of how natural-language judgments can fit inside an established enterprise interface without forcing database users into a separate chatbot.
Actian is also advertising Europe-remote openings for an AI-augmented software engineer and an AI enablement lead on its careers page. Zachi's announcement places his hire within that broader engineering push, although his public post focuses on the mission rather than an organizational title.
The sequence moved unusually fast: TypeSafe released Jev on September 15th, Zachi published the PostgreSQL extension on September 17th, and he announced the Actian move on September 21st. Actian effectively recruited from a live product demo built around a model that had been public for less than a week. For Zachi, the open-source repository became the interview.