- Building a multi-agent AI platform for restaurant operators on GCP Cloud Run with Google ADK.
- Architecting RAG pipelines over BigQuery with dbt-transformed data models.
- Designing access control systems that scope every agent call to authenticated user permissions.
- Co-authoring a public engineering blog series on the systems behind Tactix.
Sahej Maharjan
AI Engineer · Full-Stack
Summary
AI engineer building production-grade multi-agent systems and RAG pipelines on GCP. Currently shipping AI at Tactix AI — turning messy restaurant data into intelligent operator tools. 5+ years across front-end leadership, full-stack delivery, and production AI systems. Co-author on DVx engineering blog; merged open-source contributor; maker on Indigo's #2 Product Hunt launch.
Work Experience
- Continued as lead front-end engineer after Shopswap was acquired by Indigo AI.
- Owned full-stack feature delivery, architecture decisions, and production releases for React and Next.js applications serving live users.
- Collaborated with design and backend teams to resolve technical challenges across cross-functional delivery.
- Led front-end engineering for Shopswap, a toolkit for brand partnerships, giveaways, and discount sharing.
- Developed and maintained frontend architecture and coding standards; delivered production features for React and Next.js applications.
- Proactively identified and resolved technical challenges and roadblocks across cross-functional delivery until acquisition by Indigo AI.
- Worked as front-end team lead overseeing day-to-day operations of the frontend development team.
- Managed scheduling, coverage, documentation, and delivery across the Preparie platform.
- Performed front-end and mobile development for all client projects.
- Wrote 200+ JavaScript components for client websites and mobile apps.
- On-boarded and mentored a new front-end developer through regular coaching sessions and work reviews.
Writing
Speeding up Tactix Co-Pilot across the data path, tool layer, and reasoning loop — with phase timing to know which system owns the wait.
Building a conversational analyst that routes across gold marts over MCP instead of holding the warehouse in context.
Daily automation for vacation-rental gap-night upsells and pool heater management across 100+ properties.
Hybrid RBAC/ACL design, single-function Postgres auth, and why BigQuery should never hold a grant.
Open Source
- End-to-end company logo uploads: client validation, S3 signed URL flow, Zod schema extension.
- Sentry error logging and toast feedback on the Codú job creation flow.
Technical Skills
Core — could teach it
- TypeScript, Next.js/React, React Native, Electron
Proficient — used in production
- Google ADK, FastAPI, n8n, MCP tools, Node.js/NestJS
- Postgres, MongoDB, Supabase, Docker
- AWS: Lambda, ECS, S3, DynamoDB, Cognito, Route 53, Secrets Manager, SQS, SNS, EventBridge, AppSync, Amplify, CloudWatch, IAM
- GCP: Cloud Run, BigQuery
Familiar — learning or light use
- Sentry, LaunchDarkly, Terraform