Anuj Lokhande

I’m a Product Engineer who owns the full lifecycle of AI products—from architecture and development to operations. I build production-grade, event-driven systems that keep latency low and quality high under real traffic.

At CreditNirvana (a Perfios company), I’m the sole engineer for an AI loan-collection voicebot and call audit platform across 3 enterprise clients. I architected a multi-model AI pipeline using Sarvam AI (TTS/ASR via Exotel), Gemini, and Claude—reaching 87% transcription accuracy on Indic language calls.

I designed a 3-layer async concurrency model (asyncio event loop, semaphore-gated in-flight limits, asyncio.gather batch fan-out) and kept p90 latency under 500ms at 100K calls/campaign scale. I also implemented a fault-tolerant event-driven architecture with priority-separated SQS queues, per-call retries, and API response logging, deploying containerized worker/API/frontend services with Docker.

Earlier, I worked as an Analyst trading EURIBOR 3M futures, ESTR futures, and OIS swaps, and contributed quantitative analysis using OLS regression, time-series, and correlation methods. I also supported ML research at IIT Guwahati, exploring eigenvalue prediction with a PyTorch FeedForward Neural Network and benchmarked results across hyperparameter configurations.