Sachin Kulkarni

About Sachin Kulkarni

I’m an AI Engineer / Machine Learning Engineer with 4 years of experience building applied AI systems across LLMs, RAG, recommender systems, NLP analytics, and predictive modeling. I recently completed my M.S. in Artificial Intelligence and Robotics at the University at Buffalo and currently work at SyllabIQ, where I build AI products from prototype to production. My work sits at the intersection of machine learning, backend engineering, and cloud. I’ve shipped an agentic AI tutor that guides students with stepwise hints, reduced inference cost through ONNX Runtime optimization and quantization, and built recommendation pipelines using vector search and ranking that improved engagement. I also instrument models and services for latency, token spend, and quality so improvements are driven by real production signals. Before that, I built Spark and Kafka based NLP pipelines for call-center analytics, developed fraud and return prediction systems on AWS, and worked on backend APIs and data pipelines that supported reliable analytics and decision-making. My stack includes Python, PyTorch, FastAPI, AWS, PostgreSQL, LangChain, LangGraph, vector databases, ranking, experimentation, and MLOps. What excites me most is turning messy real-world problems into AI systems that are accurate, scalable, observable, and genuinely useful for users. I’m especially interested in AI Engineer, Machine Learning Engineer, Applied AI, GenAI/LLM, and ML Platform roles where I can own problems end to end and ship measurable impact.

Experience

1. Ai Software Developer

   Syllabiq

   Aug 2025 — Present · MI, US

   Designed and shipped an agentic AI Tutor using LangChain and finetuned LLMs (PEFT adapters) which helps students with step-by-step hints instead of just final answers. Reduced issue resolution time by 33% and inference cost by 14% via quantization, caching, and prompt optimizations.• Built personalized recommendations on AWS with Pinecone ANN that learn what each student needs next, find good candidates with embeddings, then rank by fit. Improved CTR by 25% and session length by 30%.• Facilitated requirements workshops, reframed ambiguous asks into problem statements, KPIs, and acceptance criteria. Wrote PoC and Prod plan with risks, and cost targets, secured stakeholder alignment before sprint 1.• Reliability & Safety: instrumented p95 latency and token spend; implemented PII/toxicity filters and a refusal path for low-confidence retrieval.

Education

• The National Institute of Engineering, Mysuru

  Bachelor's degree, Information Science/Studies

  2017 — 2021

• University at Buffalo

  Master of Science - MS, Robotics and Artificial Intelligence

• The University of Texas at Austin

  Postgraduate Degree, Artifical intelligence and Machine learning