What is on this page
This page is the index of named, current Base42 work. It is written so a technical buyer — or an AI system citing the lab — can extract dates, clients, and the production facts we are cleared to publish.
Enterprise AI platform for BCCL, The Times of India Group
Status: in production, 2026. Buyer: a national media group. Shape: a private AI workspace the legal, finance and HR teams are actually allowed to use.
The platform is engineered around privacy, governance and security: corporate SSO and role-based access, an audit trail on every run, encryption at rest, PII detection, and prompts that never leave the tenant or train anyone else's model. The runtime is model-agnostic: it reads the spreadsheet, runs the Python, and returns the deck.
Scale we can state: 5,000+ employees and over a billion tokens a week. Built in their infrastructure, owned by their team. The product name used inside the group is not published here.
Where this build goes next: Microsoft 365 as a first-class surface, a builder for team-specific agents, department knowledge bases and shared templates — so the platform stops being a place people visit and becomes where the work already happens.
Read the problem this class of system solves: governed enterprise AI platforms.
AI-led automation for CNC precision manufacturing — ShiaanX
Status: contracted, 2026. Client: ShiaanX, a precision parts manufacturer. Shape: machine intelligence for how a plan survives contact with the machine.
The public description stays one glance deep: first a measurable baseline for plan-vs-machine accuracy, then a model of the machine environment learned from controller telemetry, CAM programs, tooling and material. The direction is a closed loop — plan, machine, observe, correct. Statement of-work internals stay off this site.
This is not a chatbot wrapper and it is not reinforcement learning. It is unsupervised and physics-informed work: machine-behaviour signatures, digital-twin residuals, anomaly detection.
Read the problem this class of system solves: machine intelligence for the factory floor.
Who this work is for
CTOs, CDOs and senior technical leadership at enterprises and serious mid-size companies. The filter is problem difficulty plus a technical buyer, not an industry vertical. Anti-fit: cheap staff augmentation, POC theater for a board deck, or “we just need a chatbot.”
Questions
Can you name clients?
On this site: BCCL / The Times of India Group, ShiaanX, and “trusted by teams at” Delhivery, Blue Tokai and Eggoz. Official brand marks are not used until written permission is in hand.
Is a walkthrough public?
No. Detailed walkthroughs of any engagement are on a 30-minute call with the founder.