Hemanth Kongara


I’m Hemanth Kongara, a Senior Data Scientist specializing in Search, Recommendation Systems, and Trust & Safety ML. I build large-scale, production-ready machine learning systems that drive measurable business impact through better retrieval, ranking, and personalization.
My work focuses on vector retrieval, learning-to-rank, deep learning, and graph-based models, with strong ownership from research to deployment. I enjoy solving hard problems at the intersection of AI systems, data quality, latency optimization, and user relevance, and I’m deeply experienced in taking ML ideas all the way to real-world scale.

Experience

  • Senior Data Scientist | Apna (Sep 2024 - Present)

    Search & Recommendation – Trust and Safety

    • Leading initiatives across search and recommendation systems to improve content discovery and personalization
    • Migrated entity-based retrieval to taxonomy-aware vector retrieval, delivering 11% SL uplift, 8% TU growth, and 11% CTA improvement
    • Built self-attention fusion layers and L4 cross-encoders with hard-negative mining for ranking improvements
    • Optimized inference pipelines using ONNX, reducing latency by 50%
    • Trained and deployed XGBoost Rankers, achieving 10% NDCG@10 uplift
    • Designed fraud-risk classification models improving recall by 8%
    • Built RAG-based company–industry mapping pipelines with Airflow automation

  • Senior Data Scientist | Monster.com (Foundit.ai) (Aug 2022 - Sep 2024)


    • Developed Learn-to-Rank (LTR) models for personalized search and recruiter ranking
    • Built GNN-based cohort models, improving NDCG@5 by 13% and NDCG@10 by 17%
    • Delivered personalized job recommendations using BiVAE, increasing apply rate by 50%
    • Built SBERT-based semantic matching systems deployed with FastAPI + Docker
    • Improved notice-period and salary prediction coverage using Elastic vector search

  • Associate | Irunway (Jul 2018 - Jun 2019)

    • Developed microservices using Spring Boot to enhance system scalability
    • Automated workflows using Camunda BPM, reducing fulfillment time by 10%
    • Designed RESTful APIs, ensuring seamless communication between systems
    • Implemented CI/CD pipelines with Kubernetes & Docker for efficient deployments

Achievements

  • Reliance Foundation Scholar (Top 1 in 40 nationwide)
  • All India Rank 92 – GATE 2020

Projects

  • Graph Multi-Attention Network (GMAN) for traffic prediction
  • DiffNet++ for social recommendation
  • Super-resolution using GANs
  • YOLOv5-based classification and segmentation
  • Panoramic image mosaicing