TODO: OctaK Edge screenshot
01
OctaK Edge
Evidence-based stock verdicts for Indian equities, where AI explains but never scores.
- Equities research
- Evidence-based scoring
- AI explanations
Anurag Raina
Product-focused engineer with 5+ years shipping web, mobile, and AI-driven systems - including the LLM-powered agents inside them.
Three things I build, each tied to something already shipped.
LLM-powered features and agents built into production apps - tool use, structured outputs, and guardrails.
Shipped
Built a Copilot-powered AI agent into the Electron Test Studio at Mozark, cutting test creation by ~2 minutes per user journey.
End-to-end web and mobile products - UI, backend APIs, and the cloud infrastructure underneath.
Shipped
Architected a no-code enterprise test-automation platform that contributed to a ₹2.5 Cr+ annual ICICI Bank deal.
Pipelines, test frameworks, and internal tools that make teams faster and releases safer.
Shipped
Replaced Robot Framework/Appium with a CSV-to-Pytest pipeline - 90%+ smaller files and 5s+ faster runs.
Three products, each with a full case study behind it.
TODO: OctaK Edge screenshot
01
Evidence-based stock verdicts for Indian equities, where AI explains but never scores.
TODO: Learner Memory System screenshot
02
An AI tutor that builds an inspectable model of how you learn - and adapts to it.
TODO: ASO Audit Agent screenshot
03
Scores any App Store listing against a 10-point rubric, with cited evidence and a prioritized fix plan.
A few principles I hold to, each earned on a real system.
The interesting failures hide at the seams - between the interface, the API, and the infrastructure. I work across all three so a feature ships as one coherent thing, not three half-built pieces handed off between people.
Slow is a bug, but guessing at the cause wastes days. I profile first - which is how a mobile app went from 2.5s of navigation lag to under 1s, and a hot query dropped from 11s to about 2s.
Coverage is not bureaucracy - it is what lets you refactor without holding your breath. On a HITRUST-certified medical-device platform that meant taking unit coverage to 99% before reworking the core.
LLMs are strong at language and shaky the moment you hand them the decision. I use them for explanation, drafting, and structure, and keep the calls that actually matter deterministic and inspectable.
The better version of a system is usually smaller than the first one. Replacing a whole test framework with a CSV-to-Pytest pipeline cut file size by more than 90% - less to run, less to maintain, fewer places to break.
Five roles, from intern to Member of Technical Staff.
June 2024 - Present
Gurugram, India
Member of Technical Staff
$40M-funded AI-driven observability platform (mozark.ai).
April 2022 - May 2024
TX, USA (Remote)
Full Stack Engineer
HITRUST-certified cloud for medical devices (galendata.com); acquired by Matrix One.
Dec 2020 - March 2022
Delhi, India (Remote)
Full-Stack Developer
Sep 2020 - March 2021
Flutter Developer (Freelance)
Jan 2020 - March 2020
Gurugram, India
Software Engineer (Intern)
What I reach for, weighted toward what I use now.
Where I build in the open - side projects, experiments, and the tools behind them.
If you need an engineer who can own it end to end, from the interface to the infrastructure, let's talk.