// full-stack & ml systems engineer
Nithin Kumar
Meena
I build the parts of a product that aren't allowed to go down APIs under real traffic, computer vision pipelines that run unattended at scale, and the real-time infrastructure underneath. Based in Hyderabad, working with teams everywhere.
client → api (rest + socket.io) → ml pipeline / database
About
I started as a frontend intern shipping UI tickets. Three years later I was the contract CTO of a healthcare startup, owning the technology roadmap, and the engineer who put a computer vision model into a pipeline making 40,000+ automated decisions a day with no human in the loop. In between, I've built real-time systems with Socket.io, designed REST APIs and MongoDB schemas for production load, and trained a segmentation model over 6,000,000 iterations until it held at 99%+ accuracy. I work across the stack because production systems don't respect the boundary between frontend, backend, and ML somebody has to own the whole path from request to result.
Stack status
Frontend
stableInterfaces built to stay fast under real data volume, not just in a demo recording.
Backend
stableServices written to be read by someone debugging them at 2am, not just to pass a code review.
Data
stableSchemas and indexes built around how the data is actually queried, not just how it's modeled.
AI / ML
stableModels trained to ship into a pipeline and evaluated on production error rates, not notebook accuracy.
Infrastructure
stableThe unglamorous layer that makes everything above it reliable enough to trust.
Deploy log
- Led the development of Drivonic, coordinating 4+ engineers while shipping 20+ production features across vehicle management, CRM, and dealership operations.
- Architected high-performance backend services including OpenSearch audience indexing, Stripe payment infrastructure, and manual append pipelines integrating KBB, Black Book, and Recall data providers for real-time vehicle enrichment.
- Designed scalable REST APIs, optimized database queries, and introduced reusable backend architecture, reducing API latency by 40% and accelerating feature delivery by 35%.
- Owned technical planning, code reviews, production deployments, and cross-team collaboration, delivering 50+ releases while maintaining high code quality and platform stability.
- Owned technology strategy and execution for a digital healthcare platform, directing engineers, designers, and stakeholders to hit product milestones on schedule.
- Designed and shipped an AI resume-parsing system using NLP and transformer models, cutting manual data entry by 70% and speeding up candidate onboarding.
- Built real-time infrastructure with Node.js and Socket.io that measurably improved platform reliability and user retention.
- Set the technology stack, microservices architecture, and a 12-month engineering roadmap for projected growth.
- Designed and shipped dynamic web apps with React.js and Node.js/Express.js across multiple concurrent client projects.
- Built REST APIs and third-party integrations, cutting integration turnaround time by 30%.
- Worked closely with design and QA to refine functionality, reaching a 25% faster delivery cycle.
- Cut software defects by 40% through systematic unit testing and end-to-end coverage across web and mobile.
- Trained a production image-segmentation model on 60,000+ images over 6,000,000 iterations in PyTorch, reaching 99%+ accuracy with a 0.5–1% inference error rate.
- Deployed it to a pipeline processing 40,000+ images a day for automated quality assessment — zero manual review required.
- Built a multi-class classifier for vehicle type and shot angle that powered customized image pipelines for a high-volume car marketplace.
- Applied augmentation and bias-correction techniques to improve generalization across lighting and shooting conditions.
- Built and maintained UI across 20+ pages, contributing to a 15% lift in measured user engagement.
- Standardized components and workflows, cutting delivery timelines by 20%.
Selected builds
Zenith Studio
A real-time collaborative workspace for creative teams — concurrent multi-user sessions, live project boards, and asset sharing. WebSocket-based state sync keeps every collaborator's view consistent without a page refresh.
AutoLead
An automated lead management platform designed to streamline lead capture, organization, and follow-up workflows while giving teams a centralized view of their sales pipeline.
Prodline
A production and operations management platform built to streamline workflows, track production activity, and provide teams with a centralized view of operational processes.
RaceKraft
A modern web platform for RaceKraft, focused on delivering a fast, responsive digital experience with structured content, product presentation, and a streamlined user interface.
Track record
MERN Stack Web Developer — Zepskill
Jan 2022 – Sep 2022. Formal training in the exact stack used across every build on this page.
Algorithmic problems solved on LeetCode & GeeksforGeeks — data structures and algorithms under real constraints, not just syntax.
Open a channel
Open to full-time backend, full-stack, or ML systems engineering roles. Remote-first, happy to work across time zones.