Abhishek Ingle

About

Backend engineer with hands-on experience building low-latency, high-concurrency real-time chat on AWS (Node.js/Socket.IO/MongoDB) and improving reliability under unstable networks, contributing to ~40% user adoption growth. Also built FastAPI-based AI assistant context retrieval (RAG) APIs with embeddings/vector search, and has strong production experience in rate-limit handling, async refactors with safe rollout, and Supabase Auth/RLS optimization.

Experience

Education

Key Strengths

• Designed and scaled real-time chat system with low latency and high concurrency (AWS, Socket.IO, MongoDB)
• Improved UX reliability under network instability via heartbeats, client acks, and buffering/retransmit
• Drove measurable impact: ~40% increase in user adoption
• Built FastAPI-based context retrieval (RAG) API using embeddings and vector similarity search
• Handled third-party rate limits with token bucket + exponential backoff middleware
• Managed incremental rollout of async inference refactor using API versioning and feature flags
• Ensured data integrity with shared Pydantic schema contracts across API and workers
• Implemented DLQ to prevent request/job loss during transitions
• Implemented Supabase Auth + RLS; optimized RLS performance by denormalizing ownership data
• Improved robustness with exactly-once semantics using Redis-backed idempotency keys to prevent duplicate retries
• Owned end-to-end production feature delivery (frontend, backend, infra, CI/CD, observability)
• Built real-time streaming LLM chat UX to reduce perceived latency
• Achieved p95 latency of ~700–900ms for conversational responses in production
• Designed and implemented RAG/ETL ingestion across Google Drive/Notion/OneDrive
• Strong reliability engineering for flaky/rate-limited APIs (token bucket throttling + exponential backoff)
• AWS production operations: ECS Fargate + ALB, multi-AZ, autoscaling, CloudWatch alarms, X-Ray tracing
• Full-stack architecture design with strong validation/migrations practices (Zod middleware, Prisma transactional migrations, shared TS types)
• Next.js SSR expertise (hydration mismatch fixes; auth via HTTP-only cookies)
• Database performance optimization (execution plan analysis, composite/covering indexes, cursor-to-set-based rewrites) with 30% latency reduction
• Improved maintainability via modular shared libraries/components and reusable hooks across Android/iOS/Web; reduced regressions and improved deployment consistency (~35%)
• High-ownership decision-making in ambiguous YC-stage startup; shipped MVP ingestion layer in under 2 weeks to enable enterprise pilots