Resilient synchronization
Re-architected Go/Gin APIs and client-side retry workflows, taking observed development-stage data loss from 50–60% to zero in production validation.
Engineering leader · Product builder
I'm Rakesh Biswal. I lead teams building dependable mobile and backend platforms for education programs operating at national scale.
01 Selected case study
National FLN platform
An offline-first foundational literacy and numeracy platform built for government-backed education programs operating across low-connectivity classrooms.
React NativeRealmGo / GinS3MongoDB
OFFLINE-FIRST FIELD SYSTEM / 2022—NOW
Assessment appField capture
Local RealmOffline storage
Sync buttonUser-triggered
Go / Gin APIRealm data upload
S3 backupData files
MongoDBPersistence attempt
Production validation50–60% observed data loss → zero
My role
Engineering Lead · Architecture & delivery
Core stack
React Native · Realm · Go/Gin · S3 · MongoDB
0Kassessments delivered
0.0Kstudents supported
0Kschools reached
0languages supported
Built for difficult conditions
Hardened audio recording, autosave, crash recovery, and offline synchronization for assessment teams working with intermittent connectivity.
A smaller, faster application
Migrated to React Native's New Architecture with TurboModules and C++ integrations, reducing the Android application from more than 100 MB to 48 MB.
NIPUN Bharat delivery across Odisha, Uttar Pradesh, and Rajasthan
Automated Bitbucket and Jenkins pipelines across QA, staging, and AWS production
Roadmaps and releases aligned across product, QA, deployment, and government stakeholders
02 Selected case study
Real-time scoring infrastructure
A high-throughput evaluator workspace designed to keep scoring fast, recoverable, and traceable while thousands of browser events compete for delivery.
Full-duplex WebSocketsPersistent event queuesAudit trailsAI-assisted scoring
The operating constraint1K–2K
evaluators scoring concurrently
Every interaction must survive bursts, reconnects, and temporary delivery failures without slowing the evaluator down.
System response
Browser-first reliability with controlled server delivery
WebSocket scoring architecture Connection active
BrowserUser events
Score · rubric · evidence
Event controllerBurst detection
Isolated Send nowBurst Form queue
Delivery bufferPersistent queue
Flush after 5s inactivity
Retry enabledBrowser transportWebSocket client
Reconnect · broadcast
events→
ack / broadcast←
BackendWebSocket server
Receive · validate · acknowledge
Scoring persistenceVersioned scores · activity history · audit trail
0total assessments scored
0total subtests completed
0students represented
Autosave that handles evaluator bursts
Isolated browser events can send as single requests, while rapid activity forms a persistent queue that flushes after five seconds of inactivity, with retry handling for failed delivery.
Every scoring decision has history
Added versioned scoring workflows, audit trails, evaluator activity tracking, and scrutiny controls for stronger quality and operational oversight.
AI support inside the workflow
Integrated Sarvam, Gemini Flash, and in-house models to support evaluators, shorten scoring cycles, and reduce scorer staffing requirements.
Improvement without interruption
Improved evaluator UI and extended workflows while active scoring operations continued.
Operational outcome
Scoring workflows could evolve without interrupting active evaluation operations, while version history, scrutiny controls, and evaluator activity remained visible.
03 Venture in progress
CTO · March 2026 — Present
English confidence, built from an Odisha-first perspective.
Leading the product and engineering of an AI-powered English-learning platform—from the public website and Flutter app to the backend, content CMS, and generation infrastructure.
Next.jsReactFlutterNode.js / ExpressTypeScriptMongoDBGoogle Cloud
Visit E‑TalkProduct directionOdia → English
Personal learning paths, relatable real-life scenarios, and AI-assisted speaking practice designed around each learner's level and goals.
PRE-LAUNCHLesson generation → internal testing
138canonical topics
12onboarding goals
60–70topics per learner roadmap
7CEFR bands · A1 to B2
01 / Generation infrastructure
A content engine built for quality, cost, and recovery.
I reduced an initial spine of 631 topics to 138 reusable canonical topics that can still form unique, goal-relevant roadmaps for individual learners. The CMS coordinates batch generation, review, approval, and publishing without discarding valid work when one asset fails.
- GPT Terra for lesson generation
- GPT Image Sunburst for lesson imagery
- Granular retries preserve successful lessons and assets
- Live queue monitoring surfaces failures and repair states
Proprietary topic names and lesson content have been anonymized.Personal without generating everything twice
Roadmaps combine CEFR progression, learner goals, Odisha-specific situations, useful distractors, and topic uniqueness—sharing a controlled content pool while preserving a tailored experience.
One system across every surface
Built the launch website, Flutter application, Node/Express APIs, MongoDB data layer, payment flow, learner roadmap, question experiences, progress tracking, and the internal content platform.
Scaling toward sustainable economics
Current work is focused on infrastructure cost after launch and finding the right balance between introductory pricing, AI usage, operational cost, and sustainable revenue.
04 What I bring
I work across the product stack—from the interface people touch to the architecture, delivery systems, and engineering practices that keep it dependable.
01
Systems architecture
Designing resilient platforms from product requirements through data models, APIs, sync protocols, and observability.
02
Offline-first products
Building mobile experiences for low-connectivity environments where every saved record and retry path matters.
03
Technical leadership
Creating clarity across engineering, QA, deployment, and stakeholder teams to move complex programs forward.
05 Career ladder
March 2022 — Present
From building HTML5 learning experiences to leading architecture, delivery, and engineering teams—while remaining close to the implementation.
2025 — NOW01
Educational Initiatives
Assistant Engineering Manager
Leading engineering teams, architecture decisions, and cross-functional delivery for large-scale education platforms.
2024 — 202502
Educational Initiatives
Engineering Lead
Led the Neev rebuild and high-stakes state deployments, turning field constraints into dependable product systems.
2023 — 202403
Educational Initiatives
Lead Developer
Owned full-stack delivery across mobile, backend, quality, and release operations for offline-first assessment tools.
2022 — 202304
Educational Initiatives
HTML5 Interactive Developer
Built interactive learning experiences and established the hands-on product foundation that grew into broader mobile, backend, and platform ownership.
06 What people say
Leadership, seen from the team
03 voices
The clearest measure of leadership is the confidence, ownership, and growth it creates in others.
Working with Rakesh has been a great learning experience. He was my first senior and has always been someone I could rely on for his knowledge, guidance, and calm approach. Even during high-pressure situations and urgent field requirements, he handled things calmly and efficiently, resolving issues quickly without letting the pressure affect the quality of his work. His expertise and willingness to help have made a significant impact on my learning and growth.
Rakesh is one of the most supportive and approachable leads I've worked with. He brings strong technical expertise to the table, but what stands out most is how generously he shares that knowledge—whether it's patiently explaining a concept from the very basics or breaking down a complex task into clear, actionable pieces. He gives us the ownership to solve and grow while providing the right guidance when we need it, which makes him not just a great lead, but someone who genuinely helps the people around him become better.
Working with Rakesh was a great experience. He understands requirements quickly, and always focuses on delivering efficient and scalable solutions.