Full Stack Developer
Solo built Inora and cut API latency by 95%, bringing response times below 200ms.
Explore my professional experience and the products I've built.
Design is not just what it looks like and feels like. Design is how it works.
Ankit Jadhav
I'm Ankit, a Full Stack Software Engineer who enjoys building products that solve real problems.
My focus is on turning ideas into scalable, production ready applications with clean, maintainable code.
From intuitive user experiences to reliable backend systems, I enjoy building across the entire development process.
Solving meaningful problems, taking ownership, and continuously improving my craft are what motivate me as an engineer.
Where I've Worked
Experience
A timeline of the teams I've shipped production code with.
Full-stack Developer Intern
Spacenos·Remote
Mar 2026 – Jun 2026
4 Months
- >Designed and shipped a production subscription system with Razorpay - order lifecycle, webhook processing, and transaction management with idempotency and race condition handling.
- >Built and owned the admin dashboard for a live astrology platform, integrated third party calendar API via Cloud Functions and managed real time call billing and user data.
- >Migrated entire platform from Firebase to Supabase authentication, call logs, blog management, admin dashboard, and all data models with full API rewrites across the codebase.
Full-stack Developer Intern
AMO Foundation India·Remote
Jun 2025 – Feb 2026
9 Months
- >Reduced dashboard data transfer by over 99% and raised API throughput by 38% in 50 user load tests by moving period over period aggregation into SQL instead of returning raw transactions.
- >Optimized bulk CSV import to run 85% faster than per row requests in benchmarks, and automated onboarding through Plaid bank sync.
- >Designed an AI support agent that answers only from each customer’s own documents and escalates to a human when unsure, with tenant isolation verified by 550 automated access tests and zero data leaks.
- >Built voice cloning and text to speech on serverless GPU inference, with in browser recording, server side audio validation, and rollback on failure to keep storage and usage billing consistent.
My Skillset
Tech Stack
Projects




































Inora | SaaS Notebook Platform
- >Launched Inora, a full-stack SaaS notebook app with workspaces, search, and 22+ block types.
- >Cut API latency by ~95% (auth 3.9s to <200ms) via middleware caching and indexing, with zero layout shift and sub-second navigation.
- >Live publishing renders notebooks as interactive pages (code, tabs, media), not static exports.
- >AI feedback chat on Gemini with Groq failover, streaming, and server-side rate limiting.
- >Razorpay freemium billing with server-side plan limits; downgrades preserve all user data.
































Cortex Stream | AI Companion
- >Built Stripe billing with durable webhooks, transactional updates, idempotency, and reconciliation; tested 10,000 duplicate and out of order events.
- >Blocked cross user access and free tier bypasses with query level authorization and atomic quotas; recorded zero unauthorized mutations across 1,000 concurrent tests.
- >Secured AI conversation memory with isolated Redis keys, ordered messages, idempotent requests, and token aware context; verified zero cross session leaks or lost messages.
- >Built Gemini and Groq failover with quota aware routing, bounded retries, circuit breakers, and latency aware fallback; targeted 95% success within 10 seconds.
- >Tracked per user AI costs using token usage, latency, and request outcomes; targeted 95% cost attribution and anomaly detection within 60 seconds.
































DevSuite AI | Multimodal Workspace
- >Built atomic AI quota enforcement across five endpoints, preventing overspending across 1,000 concurrent requests.
- >Designed recoverable AI job processing with persistent states, retries, and provider reconciliation; targeted 90% recovery under simulated failures.
- >Built provider failover with quota aware routing and circuit breakers, targeting 30% fewer failed generations.
- >Enforced per user AI budgets with usage based cost tracking and spending limits; validated zero budget overshoots.
- >Implemented end to end generation tracing with correlation IDs, targeting 100% state transition traceability.
































SpendWise | Personal Finance Platform
- >Built atomic CSV imports with validation and deduplication; tested 1,000 failure scenarios with zero duplicates or partial commits.
- >Built an auditable financial ledger with precise currency handling; verified zero calculation discrepancies.
- >Enforced account isolation with user scoped queries and server authorization; recorded zero unauthorized operations across 1,000 security tests.
- >Optimized reporting on 100,000 transactions, achieving 97% lower p95 query latency.
- >Built recoverable imports with batch tracking and retry safe processing to account for every transaction after failures.
Designing and building modern applications.
Open to meaningful collaborations and opportunities.