AI Engineer · Engineering Leader · Startup Founder
14 years of software engineering. Currently co-founding MaidLink — a real business with real customers, and I'm building the technology to scale it.
My husband got laid off in 2023. Instead of job hunting, he went out and built a real cleaning business by hand — recruited maids, trained them, found clients, and ran actual jobs. No app. No platform. Just hustle and a clear vision.
I got laid off in February 2026. So I went all-in on building the technology to scale what he'd already proven works.
That's MaidLink. And that's what building for real looks like.
Before this, I spent 14 years as a Software Engineering Manager and Tech Lead — shipping products, building teams, and never quite letting go of the code. I'm the kind of engineering leader who still writes it. That combination is rarer than it should be.
Production systems, not slideware. Real constraints, real numbers.
OpenAI API on AWS Lambda, scoring real-estate acquisition targets in parallel with an automated valuation model. Nobody asked for this — I built it in a week and showed the business team the output directly.
Inherited an abandoned, AI-generated document platform with zero handover. Rebuilt it into a VPC-only system indexing 200K+ real estate documents, metadata-first search via OpenSearch, RBAC via Auth0. Replaced Box entirely, saving $50K–75K/year.
Amazon Bedrock + Nova Lite analyzes uploaded room photos and returns time estimates, cleaner-count recommendations, and a full checklist. $0.0012 per analysis.
Internal agent system coordinating estimator tasks. Evaluated four LLM providers on cost, privacy, and deployability before choosing Groq over the more familiar options.
Two-sided marketplace for professional cleaning services · Calgary, CA
The business exists and the app is live — real maids, real clients, real jobs. People are actively using it.
The flagship feature is an AI Estimator: a client uploads photos of their space directly to S3 (Lambda never touches the bytes), and Amazon Nova Lite on Bedrock returns room-by-room time estimates, 1-cleaner and 2-cleaner hours, a full cleaning checklist, and coverage warnings if photos are incomplete.
Cost per analysis: $0.0012. At 1,000 analyses/month: $1.20. Built on AWS Bedrock + Amazon Nova Lite on Lambda.
I document the build publicly on LinkedIn — the real decisions, the tradeoffs, the mistakes.
Follow the build log →Production work. Everything here is live or in active use.
Full-stack agentic AI platform for MaidLink. AI Estimator uses AWS Bedrock + Amazon Nova Lite to analyze room photos and return time estimates at $0.0012/analysis. Sole engineer, live in production.
Agentic pipeline that scans RSS news feeds, evaluates signals with Claude on Bedrock, drafts LinkedIn posts, and emails for approval before committing to a content repo. Fully automated, zero manual curation.
AI-powered library management system on AWS serverless. Members borrow, hold, and get AI book recommendations via Amazon Nova. Librarians get an admin dashboard with AI loan analytics (Claude Haiku), scheduled background agents for overdue detection and demand signals, and a human-approval queue before any consequential action fires.
Construction-compliance prototype built for a Founding Product Engineer take-home assessment (inspired by Tough Leaf's ClearComply concept). Compliance managers track subcontractor certificates of insurance, workers' comp, and EMR expiration; AWS Bedrock (Claude Haiku) extracts structured data from uploaded documents. Two roles: compliance manager and subcontractor portal.
Demo (synthetic data) — Compliance Manager: compliance / ClearComply2026! at /login · Subcontractor Portal: apex-electrical / Passw0rd! at /sub-portal/login
Most production work lives in private repos. Public repos and contributions at github.com/sindhujaIBM.
View GitHub →From Avenue One's deal-lifecycle rebuild — the kind of detail that only shows up when you actually built the thing.
AI wasn't on the roadmap when this shipped
Avenue One needed a faster way to know which cold-called sellers were worth pursuing. Nobody asked for an AI solution — I built a working version in a week and showed the business team the output directly.
Shipped to production on OpenAI. Now part of how every deal gets sourced.
No-code platform → serverless AWS
Avenue One ran its entire deal lifecycle — sourcing through closing — on Unqork, an expensive, inflexible no-code platform. I owned re-architecting the two core systems onto AWS, arriving with no serverless background and a 60-day deadline.
OMS shipped in 70 days against a 60-day target. Combined with Transaction Manager, eliminating Unqork saved Avenue One roughly $650K/year.
An abandoned AI-built codebase, handed over cold
An engineer built the first version of DocHub with Claude, then left before anyone could run it — no documentation, no ERDs, no deployment notes. My manager handed it to me, offhand, in a meeting. I dug in, understood it, and rebuilt it.
Shipped over 4.5 months, weekly stakeholder demos throughout. Replaced Box entirely — $50K–75K/year saved, 200K+ documents indexed, zero public endpoints. At the final demo: "This is excellent." Shipped just before I was laid off in February 2026.
I've spent 14 years in the room when things ship and when they don't. I've built teams, set technical direction, managed delivery under pressure, and made the calls that don't have clean answers.
The part that makes me unusual: I never stopped coding. Most engineering managers drift away from the code as they move into leadership. I stayed close to it by choice — because it makes me a better leader, a better reviewer, and a better judge of what's actually hard vs. what just sounds hard.
Right now I'm building a production AWS serverless application on Bedrock — solo, while co-running a startup. Shipping real agentic AI, not just talking about it. That's the proof.
The rare part isn't any one piece of this on its own — it's the combination. I can architect the system, ship the AI feature, lead the team through it, and still explain the trade-off to a stakeholder in plain language.
I write about what I'm actually building and learning — no manufactured inspiration, no productivity clichés.
Currently open to Staff AI Engineer, Applied AI Engineer, Forward Deployed Engineer, and Engineering Manager (AI) roles — remote, Canada or US. If what you've read here matches what you're looking for, reach out directly.
Building MaidLink? Interested in what we're doing in Calgary? Check out MaidLink →