Syed Omer Shah

- Open-source contributor for productionized lab workflows, reducing failure cases and improving project completion consistency. - Achieved a resolution rate of 92%+ while mentoring 7 cohorts and supporting 280+ learners through the forum.

Led the design and optimization of AI-driven pipelines and research projects in the talent acquisition domain, focusing on resume parsing, candidate ranking, and automation. Key contributions include: Developed a hybrid GPT-4 mini + Mistral OCR pipeline to improve resume parsing accuracy, reducing latency from 15s to 3.3s per resume (7.3s for full accuracy) through parallel processing and multi-threading. Built modular AI agents for parsing, ranking, scoring, report generation, and outreach, optimizing token usage and chain-of-thought prompts for efficiency. Implemented a custom GPT-based ranking pipeline and integrated Affinda API for candidate scoring and transparent justification, improving stakeholder trust. Designed RESTful integrations with CATSone API and Python ETL scripts for real-time candidate sourcing, automating data fetching, normalization, and storage in MongoDB, reducing manual data wrangling by 80%. Explored AI agent frameworks (Hugging Face Smol Agents, Crewai) and micro-agent patterns, prototyping workflows on AWS Bedrock for OCR and other AI tasks. Optimized API calls and batching, achieving sub-second scoring latency at scale and streamlining end-to-end AI-driven recruitment workflows.

Contributed to building advanced AI-driven systems and enterprise promptware solutions, focusing on knowledge-based reasoning, agentic AI, and generative AI applications. Key achievements include: Developed a graph RAG engine for spiritual AI, integrating a knowledge graph and iterative query models to reason over texts like Holy Scriptures, enabling complex moral reasoning tasks. Designed and implemented AI knowledge structures, query models, and reasoning pipelines, optimising retrieval and inference performance for large-scale text datasets. Conducted prompt engineering, hyperparameter tuning, and fine-tuning experiments to maximise accuracy and relevance of AI outputs, achieving high-quality reasoning with minimal latency. Built modular AI agents for tasks including document ingestion, query handling, and reasoning output generation, improving system scalability and maintainability. Collaborated on enterprise-grade AI services for US clients, integrating generative AI models into practical workflows with measurable improvements in reasoning speed and precision. Delivered solutions that enhanced retrieval and response accuracy by over 90%, demonstrating the effectiveness of structured AI reasoning over unstructured texts.

Passionate about driving online presence and impactful recruitment, I successfully increased TechSATH's LinkedIn page followers by a staggering 2000%. My internship involved: Crafting engaging recruitment materials that attracted 5 times more applicants. Optimizing website SEO and managing social media platforms for brand awareness. Developing content strategies and analyzing marketing data for insightful decision-making. Scheduling and coordinating interviews for future intern success. Results: 2000% increase in LinkedIn page followers. 500% increase in candidate applications. Enhanced online presence and brand awareness. Streamlined recruitment processes for smooth onboarding.

Edited and Compiled the 2020 Annual day album for LFHS, which was published online