Open to opportunities

Ashraf User

@ashrafuser3

Principal Engineer specializing in AI/data/backend systems—building production LLM platforms and scalable web/game products.

Pakistan

Message

What I'm looking for

I want to own complex AI/data/backend systems end-to-end—shipping production-quality products, improving reliability with SLOs, and building fast feedback loops from telemetry, experiments, and measurable outcomes.

I’m a Principal Engineer with 10 years of end-to-end product ownership, from architecting petabyte-scale data pipelines and training production LLMs to shipping Unity multiplayer games and pixel-perfect React frontends. I’m equally at home writing a Rust microservice, fine-tuning transformers, or designing real-time game physics—because I value owning the problem end to end. Hiring managers trust me for the rare ability to own any layer of the stack on day one.

At Devsinc, I designed and deployed a production RAG platform serving 4M+ daily queries with an LLaMA-3 70B backend and a FastAPI inference layer, cutting hallucinations by 61% versus a GPT-4 baseline. I architected a multi-tenant AWS data lake ingesting 2TB/day via Kafka (S3 + Delta Lake + dbt + Snowflake) and reduced query latency from 45s to 900ms using a ClickHouse OLAP layer. I also built a Next.js 14 + tRPC analytics dashboard for 600 enterprise customers, introduced feature-flagged A/B testing (Statsig), and implemented event-driven telemetry (Kafka → Flink → BigQuery).

Earlier, I was the sole data engineer at LemonBox, building the platform from scratch with Airflow on EKS and dbt models, and deploying fraud detection that reduced false positives by 42% while saving $3.2M/year. I’ve also re-architected backend systems into FastAPI microservices on Kubernetes, built GraphQL APIs that eliminated N+1 queries, and shipped multiplayer netcode and performance-focused graphics work in Unity—so I bring both engineering depth and product practicality to every release.

Experience

Work history, roles, and key accomplishments

Current

Principal Engineer AI/Data

Current

Devsinc

Jan 2022 - Present (4 years 6 months)

Designed and deployed a production RAG platform serving 4M+ daily queries, cutting hallucination rate by 61% versus a GPT-4 baseline. Architected an AWS multi-tenant data lake ingesting 2TB/day and reduced query latency from 45s to 900ms using a ClickHouse OLAP layer.

RAG fastAPI Pinecone AWS S3 Delta Lake Kafka DBT Clickhouse Next.js

Senior Data & ML Engineer

LemonBox

Mar 2019 - Dec 2021 (2 years 9 months)

Built the founding data platform from scratch (Airflow on EKS, dbt models, Redshift), supporting end-to-end analytics for a 50-person org. Trained and served fraud-detection models (XGBoost + PyTorch), reducing false positives 42% and saving $3.2M/year.

Airflow DBT Redshift XGBoost PyTorch Sagemaker Kafka Streams Redis Metabase

Software Engineer, Backend & Platform

Apozy

Jun 2017 - Feb 2019 (1 year 8 months)

Re-architected a Django monolith into 12 FastAPI microservices on Kubernetes, reducing cold-start time from 14s to 0.8s and enabling 30x horizontal scale. Built a GraphQL API gateway that eliminated N+1 queries and reduced average API response time from 340ms to 47ms.

Django fastAPI Kubernetes PostgreSQL Citus GraphQL Strawberry Celery Redis

Game Developer & Python Engineer

MindForged Studios

Aug 2015 - May 2017 (1 year 9 months)

Shipped a Unity co-op shooter and built authoritative multiplayer netcode with client-side prediction in Mirror, maintaining under 60ms perceived latency at 100 CCU. Authored an HLSL lit shader pipeline and reduced draw calls by 58% using GPU instancing and static batching.

Unity Mirroring HLSL GPU Instancing C Netcode For GameObjects

Junior Software Engineer (Python)

Systems Ltd

Jul 2014 - Jul 2015 (1 year)

Wrote Python ETL pipelines to ingest 20+ third-party REST/SOAP APIs into PostgreSQL and automated scheduling with Celery Beat. Built a Flask dashboard for real-time pipeline monitoring and delivered a production project within 3 weeks.

Python ETL Pandas SQLAlchemy Flask REST APIs PostgreSQL Data Ingestion