ML Engineer
Talrop · Mar 2025 – May 2026
- Built and deployed an LLM-driven recommendation system that learns user preferences from likes, comments, and shares, improving feed relevance and user engagement across the platform.
- Designed and implemented a RAG-based AI tutoring system using LangChain and vector search, enhancing learner experience through personalized and curriculum-aligned responses.
- Reduced data acquisition costs by generating diverse synthetic training datasets using DeepSeek V3 API, accelerating model iteration cycles while maintaining training quality.
- Trained and optimized BERT and LSTM models on real and synthetic data, achieving ~87% accuracy and enabling reliable production inference.
- Productionized ML models on AWS SageMaker with automated retraining and scalable endpoints, improving system reliability and reducing manual ML operations.
- Implemented event-driven ML pipelines using AWS EventBridge to connect real-time application events with inference workflows, lowering latency and operational bottlenecks.
- Minimized LLM hallucinations by integrating ChromaDB-backed retrieval and traceable response grounding, increasing trust and consistency in AI outputs.
- Enabled multimodal AI interactions (text + image inputs) to support richer learning workflows, expanding platform capability and accessibility.
- Authored technical documentation covering ML architecture, APIs, and lifecycle workflows, improving CI/CD adoption and cross-team onboarding efficiency.

