Krishna Yalamanchili

About Krishna Yalamanchili

As Principal Machine Learning Engineer at Altana, I specialize in designing advanced multilingual and multimodal semantic search systems. My work includes leading the development of an in-house BERT-based facility matcher and agentic shipment enrichment pipelines using LangGraph and ReAct frameworks. These initiatives have advanced global supply chain intelligence through innovations in data enrichment and classification. Leveraging expertise in AI agents, large language models, and Databricks, I am motivated to tackle complex challenges in semantic search and data quality at scale. My mission is to empower businesses with scalable, AI-driven solutions that enhance operational efficiency and decision-making.

Experience

1. Principal Machine Learning Engineer

   Altana

   May 2024 — Present · San Francisco, CA, US

   Next-gen facility matcher: Initially built a hybrid BERT + LLM reranker that fused name, address, HS-code, graph-proximity, and shipment features to lift facility-matching precision/recall 13–15 %, then superseded it with a lean in-house BERT cross-encoder trained on 1.2 M curated pairs, nudging automatic matching recall to 90% without precision loss and delivering a 32× semantic-matching speed-up (8 s → 0.25s). Optimized OpenSearch queries with custom English and non‑English analyzers and country detection.• Agentic shipment enrichment: Designed a LangGraph / ReAct pipeline that plans, executes, and evaluates RAG calls to enrich importer/exporter data and goods descriptions with HS‑level attributes—groundwork for automated labeling and multi‑stage HS‑code classification.• Facility discovery & deduplication: Replaced the legacy matcher with a fine‑tuned BERT vector index plus OpenSearch rules, boosting recall and enabling facility‑grouping logic that reveals corporate hierarchies and removes duplicates.• Architecture & standards leadership: Adopted the full Model Context Protocol (roles, function calls, long contexts) ahead of industry rollout and now spearhead an in‑house BERT search model and multi‑stage HS classifier to power next‑gen supply‑chain intelligence.

Education

• Jawaharlal Nehru Technological University

  Bachelor's Degree, Electrical and Electronics Engineering

  2003 — 2007

• Northeastern University

  Master's Degree, Computer Engineering

  2007 — 2009

Skills

• Java Enterprise Edition
• Intellij Idea
• Java
• Shell Scripting
• Javascript
• Jasmine Framework
• Angularjs
• Akka
• Software Development
• Restful Architecture
• Bootstrap
• Elasticsearch
• Localization
• Ionic
• Gruntjs
• Pl/Sql
• Rest
• Css
• Html
• Node.js
• Mysql
• Distributed Systems
• Scala
• Spring Framework
• Representational State Transfer (Rest)
• Cascading Style Sheets (Css)
• Html5
• Sql
• Restful Webservices
• Linux
• Hibernate
• Bash
• Eclipse
• Unix
• Multithreaded Development
• Apache Spark