Syed Abdul Khader

Senior AI Engineer | Agentic Systems | Evals

Handle

@syedak

Role

Senior AI Engineer

Company

Amuse Labs

Location

Bangalore

About

I design and build production multi-agent AI systems using Google ADK, MCP, LLMs and VLMs. I develop RAG and evaluation workflows, deploy FastAPI services on GCP, and improve AI-assisted creation and retrieval performance.

Experience

• Senior AI Engineer at Amuse Labs
  May 2024 to Apr 2026 - Bangalore, India
  • Designed and built PuzzleMe Agent, a production Google ADK/MCP multi-agent system spanning LLMs/VLMs, with planning, memory, structured outputs, and tools for Google Search, Wikimedia, and internal APIs. • Architected pgvector RAG for the PuzzleMe Agent with Gemini embeddings on Vertex AI; benchmarked HNSW, IVFFlat, and DiskANN vector indexes, selecting DiskANN for fastest retrieval; added BM25, HyDE, and Cohere Reranker. • Increased creator output volume by 60% through AI-assisted creation workflows. • Built Promptfoo eval loops with SME-calibrated LLM-as-Judge on 2-5% sampled production traces, heuristic evals, golden-dataset promotion, and CI gates; tracked tool accuracy, failures, latency, context, tokens, and session cost. • Deployed FastAPI AI services on GCP with Docker Compose and GitHub Actions, handling 15K-20K daily requests, including 10K agent requests, with 2x higher concurrency.
• Research Scholar at Mass General Hospital + Harvard University
  Apr 2023 to Apr 2024 - Boston, USA
  • Achieved SOTA in surgical video segmentation (92.26% binary, 74.43% multi-class IoU) by improving video object segmentation pipelines, benchmarking strategy, and reliability-focused evaluation. • Built an annotation platform integrating SAM-Med2D and XMem++; resident and medical-student annotation with surgeon review reduced labeling time by 75%. • Benchmarked SAM/SAM2, XMem/XMem++, YOLO, and Mask R-CNN on Lambda Labs GPUs across dataset sizes and augmentation strategies, using ablation studies to quantify their impact on performance.
• Data Scientist at Infinstor
  Jun 2021 to Jul 2022 - Remote
  • Built a LayoutLM-based document AI pipeline with Tesseract, PaddleOCR, handwritten OCR, and YOLO field detection for financial and tax forms, reducing processing time by 90%. • Implemented transformer-based anomaly detection for SQL queries and AWS CloudWatch logs using LogBERT, with automated metrics, drift monitoring, and error-pattern analysis.
• Dev Business Analyst at Infor
  Jul 2019 to Apr 2021 - Hyderabad, India
  Engineered SQL ETL and mapping scripts for 4 enterprise CRM/ERP migrations spanning hundreds to thousands of tables from Microsoft Dynamics CRM and Salesforce into Infor.

Education

• Plaksha University
  PGD, Artificial Intelligence
  2022 to 2023
  Technology Leaders Program. Credit standing: Gold Medalist (GPA: 9.6). Awards: Best student committee and Spirit of Plaksha. Teaching Assistant - NLP w/ Prof. Monojit Choudhury
• JNTUH College of Engineering Hyderabad
  B.Tech + MBA, Mechanical Engineering
  2014 to 2019

Skills

• Python
• SQL
• FastAPI
• Postgres
• pgvector
• GCP
• Vertex AI
• Docker
• Google ADK
• MCP
• Multi-Agent Workflows
• Tool Calling

Links

• Website
• LinkedIn

This profile as JSON