maaz.career / status running
Maaz Ashraf
Lead Backend Engineer. Search, routing, onboarding, platform infrastructure.
Lead Backend Engineer at Revcloud / Alyson.ai. Owns search over 800M+ identities, rebuilt the request router from 1500ms to under 300ms, and shipped enrichment work tied to $300K+ in additional quarterly revenue. Works mostly in Node.js, TypeScript, Python, AWS and Elasticsearch.
I like the part of a product nobody sees until it gets slow. Most of my work is making a request go somewhere, find something, and come back before anyone notices it left.
- 800M+ identities, <500ms search
- 1500 → <300ms request router
- +30% lead submissions
- $300K+ extra revenue per quarter
start here → hiring? the trace and the resume. engineer? the design notes. just curious? fire a request.
01 trace the whole career as one request
click a span · ↑ ↓ move · ← → collapse / expand
02 fire a request the router, before and after
The intelligent router sits on the hot path of every request. I rebuilt it on Lambda, ElastiCache, EventBridge and CloudWatch. Send it some traffic and watch both versions at real speed.
Nothing fired yet.
Simulated. Each latency is sampled around the measured numbers (about 1500ms before, under 300ms after) and played back in real time. More in MA-008.
03 design notes numbered in the order they happened
One note per system, newest first. Dates inside a role are approximate.
-
MA-0122023–nowRelease pipelines you can roll back
- where
- Revcloud / Alyson.ai
- built
- CI/CD pipelines and infrastructure automation with CloudFormation and GitHub Actions.
- result
- More reliable releases and safer rollbacks.
- stack
- CloudFormation, GitHub Actions, AWS
-
MA-0112025An Athena-backed data access layer
- where
- Revcloud / Alyson.ai
- built
- Low-latency querying and dynamic dataset generation over structured data in S3.
- used by
- Product and reporting workflows.
- stack
- Athena, S3
-
MA-0102025People and geo-location search−70% latency
- where
- Revcloud / Alyson.ai
- built
- A large-scale search engine indexing 80M+ people and 100M+ geo-locations.
- result
- Search latency down 70%. Contributed to a 20% lift in conversions.
- stack
- Elasticsearch
-
MA-0092024–25AI-assisted CSV onboarding and schema mapping30+ connectors
- where
- Revcloud / Alyson.ai
- built
- Led the platform: CSV ingestion, AI-assisted schema mapping, and automated connectors across Google Ads, Google Drive, Meta, Salesforce and FreshSales.
- result
- 30+ connectors shipped. Manual onboarding work down 70%.
-
MA-0082024Rebuilding the intelligent router1500 → <300ms
- where
- Revcloud / Alyson.ai
- problem
- A mission-critical router on the request path took around 1500ms to execute.
- built
- Rearchitected it on Lambda, ElastiCache, EventBridge and CloudWatch.
- result
- Execution time under 300ms, an 80%+ cut. Race both versions in 02.
- stack
- Lambda, ElastiCache, EventBridge, CloudWatch
-
MA-0072023–25Search across 800M+ identities<500ms
- where
- Revcloud / Alyson.ai
- built
- Owned the search architecture for a platform managing 800M+ identities, with Elasticsearch-based retrieval.
- result
- Sub-500ms responses at that scale.
- stack
- Elasticsearch, Node.js, AWS
-
MA-0062023–24Profile enrichment on the identity graph+$300K / quarter
- where
- Revcloud / Alyson.ai
- built
- A system on top of the company identity graph that detects unconverted users and enriches their profiles.
- result
- +30% lead submissions, +20% conversion rate, +15% revenue per click, and $300K+ in additional quarterly revenue.
-
MA-0052023MongoDB aggregation tuning50%+ faster
- where
- Nextbridge
- built
- Reworked aggregation pipelines over large user and transaction datasets, and raised automated test coverage with Jest and SuperTest.
- result
- Aggregations more than 50% faster. Test coverage up 25%.
- stack
- MongoDB, Jest, SuperTest
-
MA-0042022–23A scheduling engine and an attendance bot
- where
- Nextbridge
- built
- A Calendly-style scheduling system and an attendance automation bot.
- result
- Less operational friction for hybrid and remote teams.
-
MA-0032022–23An enterprise HR automation platform
- where
- Nextbridge
- built
- Led the backend end to end: resume filtering, candidate tracking, interview scheduling and internal coordination. Modular APIs and service workflows for a multi-portal SaaS, integrated with Twilio, SendGrid, Stripe, HubSpot and Webflow.
- shipped on
- AWS (EC2, S3, Lambda, CloudWatch), with CI/CD automated in GitHub Actions.
- stack
- Node.js, NestJS, MongoDB, AWS
-
MA-0022021–22Real-time chat and a support bot for Swvl
- where
- TDC (The Dev Corporate)
- built
- Chatbot and real-time communication services for the Swvl platform.
- result
- Faster support and more automated workflows.
-
MA-0012021–22Rewards and redemption for Peoppl
- where
- TDC (The Dev Corporate)
- built
- Core modules for an employee engagement platform: gifting, an internal e-store, digital certificates, inventory and notifications, plus transactional workflows for reward redemption.
04 facts some of these are true
- His p99 is other people's p50.
- Believes every slow endpoint has a cache it hasn't met yet.
- Has never met a CSV whose headers he couldn't map. Has met several he wishes he hadn't.
- Once cut a router's execution time by 80%. Has been insufferable about it ever since (see 02).
- Reads CloudWatch dashboards for fun.
05 errata after Knuth
If you find a real technical error in any design note, email me. The first person to report each one gets a coffee on me, and their name goes here.
found so far: 0
06 standing invitation
I'm always glad to talk about search at scale, latency budgets, event-driven systems on AWS, and getting messy data into clean schemas.
Email is best. Tell me what you're building and what's slow.
- itsmaazashraf@gmail.com
- linkedin.com/in/maaz-ashraf
- github
- github.com/itsmaazashraf
- resume