Career | Waleed Shoaib

Professional Career

Aeyron — Lewes, DE & Islamabad | Apr 2026 – Present

Problem: Enterprise clients required deterministic guarantees from probabilistic LLM outputs before deploying to high-stakes healthcare and finance environments.
Solution: Architected an evaluator-first roadmap for 3 production AI products across MENA and USA, ensuring strict structural scaffolding (RAG, validation layers, semantic routing) over raw LLM generation.
Engineered real-time YOLO-based computer vision pipelines—handling the full lifecycle from annotation strategy to sub-50ms low-latency inference at the edge.
Implemented granular telemetry using Prometheus and Grafana, ensuring observability across the entire inference graph.

Aeyron — Islamabad | May 2025 – Apr 2026

Problem: Single-prompt LLM solutions were hallucinating on complex, multi-step forex data and manual analysis was bottlenecked.
Solution: Built autonomous, multi-agent collaborative systems using LlamaIndex and ReAct, improving financial analysis accuracy by 35%.
Designed a high-throughput forex intelligence pipeline ingesting raw market data into ClickHouse with sub-500ms latency.

Catalic AI — Remote | Mar – May 2025

Problem: Ingesting massive FOMC datasets caused memory exhaustion and slow contextual retrieval for research agents.
Solution: Built robust ETL pipelines and established long-term semantic memory architectures via Milvus, reducing processing time by 10% and scaling retrieval.

Objex — Montreal, Canada | Mar 2024 – Feb 2025

Problem: Legacy search architecture yielded poor semantic relevancy and severe API latency spikes.
Solution: Migrated to a cloud-native microservices platform on GCP, integrating RAG components to drive a 40% improvement in search relevancy.
Optimized core GraphQL APIs (achieving a 25% latency reduction) and decoupled services using event-driven Pub/Sub pipelines.

Eastlancer — Islamabad | Mar 2022 – Feb 2024

Problem: Monolithic backend structures made it extremely difficult and brittle to integrate emerging third-party AI services.
Solution: Built scalable, decoupled backend APIs with Node.js/Express, ensuring resilient API integrations without cascading timeout failures.

Global | 2024 – Present

Problem: Early-stage startups were building brittle LLM prototypes based entirely on prompting, failing in production.
Solution: Advised technical founders on transitioning to robust, evaluator-first architectures. Architected custom RAG pipelines for legal and compliance firms requiring 100% deterministic citation guarantees.