Open to opportunities

Shashwat Mallick

@shashwatmallick

Message

AI/ML engineer focused on low-latency LLM and speech systems.

India

Message

What I'm looking for

I seek hands-on roles building production LLM and speech systems with emphasis on low-latency inference, reliability, and deployable pipelines in collaborative, engineering-driven teams.

I am an AI/ML engineer specializing in low-latency speech-to-LLM pipelines and efficient on-device inference. I combine systems engineering with applied NLP to build production-ready, fault-tolerant AI services.

In recent internships I designed streaming STT→LLM systems, integrated multi-agent LLM workflows using LangChain/LangGraph, and implemented quantized-model inference that cut compute load and reduced latency by 40–50%.

I have built RAG systems and local-LLM deployments optimized with GPTQ/4-bit and LLM.int8 quantization, and engineered retrieval stacks using FAISS and hybrid TF-IDF approaches to improve evidence ranking and search latency.

I focus on observability and reliability—raising uptime to 99.9% through inference-level optimizations, monitoring, and fault-tolerant serving—and I deliver reproducible deployments via Docker and modern web frontends with Next.js.

Experience

Work history, roles, and key accomplishments

AI Systems Intern

FiXitAI

Jun 2024 - Dec 2024 (6 months)

Built an end-to-end speech→LLM streaming pipeline with sub-second latency, designed a low-latency quantized-model feedback engine reducing per-request latency 40–50%, and improved system reliability to 99.9% through inference optimizations and fault-tolerant serving.

LangGraph Low Latency Inference Serving Caching Monitoring Fault Tolerance

Python / ML Intern

AgentStat

Sep 2023 - Jan 2024 (4 months)

Integrated ML APIs (e.g., Whisper) into production workflows and built a TensorFlow computer vision pipeline, profiling and applying model-level optimizations to accelerate on-device inference.

Python TensorFlow Whisper Profiling Computer Vision Integrated Production