I'm a backend engineer at Deloitte building async data pipelines that process 20M+ records/day on GCP, with a 35% API latency improvement shipped through N+1 elimination and algorithmic refactoring. My stack is Python — Django, FastAPI, DRF — with Celery, Redis, PostgreSQL, and Cloud Run for the infrastructure layer.
Outside of Deloitte, I've published 100+ technical articles at Unite.AI covering LLMs, multimodal models, and AI infrastructure — and I maintain two open-source backend tools, APIEngine and MeshEngine, both built to production standards with observability, automated tests, and real deployment pipelines.
I care about systems that are observable, maintainable, and documented well enough that the next engineer doesn't have to guess. If you need backend depth and the communication skills to match — that's the combination I bring.
Backend engineer on ChangeScout — an enterprise OCM platform serving 150+ global clients within a 40-person eng/PM/QA team.
Contributing writer covering cutting-edge AI/ML research for a global developer and research audience.
Backend internship focused on productionizing computer vision model inference.
Across distributed systems, backend engineering, and AI/ML communication
“Kunal stood out as a strong backend engineer with a sharp eye for performance. He built high-throughput APIs and data pipelines at scale — consistently focused on reliability and optimization. Proactive, collaborative, and takes full ownership.”
“Kunal brings a genuine problem-solving mindset to complex backend challenges. What stood out most was his ability to bridge development and QA — resolving critical issues during SIT cycles with clarity and ownership.”
“Reliable, skilled, and always willing to step up. Kunal has a strong grasp of backend fundamentals and delivers high-quality work with real ownership — consistently.”
“Kunal translates complex AI topics — agents, multimodal systems, modern ML — into content that resonates with both technical and broader audiences. He brings a thoughtful, analytical approach that's rare in the AI writing space.”
Mock APIs and localhost servers don't behave like production systems — no real validation, no persistence, no network behavior. Integration issues surface late in development when changes are costly.
Open-source CRM API sandbox platform. Developers get a complete REST backend with 14 CRM objects, JWT authentication, rate limiting, and full request logging — instantly accessible without any setup. Built with a custom URL dispatcher, per-user namespace isolation, Celery + Redis background jobs, and full CRUD with filtering, sorting, and pagination.
250K+ API calls/user/month. Full request logging, JWT auth, and configurable rate limits out of the box.
Distributed systems research and defence/disaster-recovery planning require testing self-healing drone mesh networks, but existing tools are either too RF-specific or too general — no async event model, no HTTP interface, no real failure simulation.
Built a backend simulation platform with a control-plane/worker-plane split: FastAPI control plane handles topology and failure-aware Dijkstra routing [O((V+E) log V)], while independent async workers subscribe to Redis Pub/Sub events for fan-out execution. Chose Redis Pub/Sub over Kafka for lower operational overhead given PostgreSQL-backed persistence. Deployed on GCP Cloud Run with Cloud SQL and Memorystore.
End-to-end simulation of self-healing mesh networks: real-time failure/recovery events, multi-hop Dijkstra routing, per-message hop-by-hop latency breakdown, and live WebSocket event stream. 25 automated tests (unit + topology).
Deep dives into backend engineering, system design, AI, and developer tooling — written for practitioners, optimized for search.
Selected articles from Unite.AI and getint.io — covering AI/ML research, distributed systems, and developer tooling.
Have a project in mind or want to discuss a collaboration? Drop me a message — I typically respond within 24 hours.