Open to opportunities

Ammar Malik

@ammarmalik1

Message

ML/AI engineer and team lead shipping production LLM and voice systems on constrained on-prem GPUs.

Pakistan

Message

What I'm looking for

I’m looking to lead production ML/AI that ships—especially LLM/RAG and real-time voice pipelines—on constrained on-prem GPUs, with strong engineering rigor in MLOps, evaluation, and reliability.

I’m an ML/AI engineer and team lead specializing in shipping production LLM systems on constrained, on-premises GPU hardware. After relocating from the US to Pakistan in 2025, I rebuilt momentum quickly—becoming the sole ML engineer at a startup and architecting its entire production AI platform solo before being promoted to lead the function.

At TekkDev, I built and deployed a production LLM platform using FastAPI, Docker, and PostgreSQL on a single RTX 6000 (24 GB). I serve Mistral 7B AWQ via vLLM with continuous batching and focus on real-world reliability under tight hardware limits.

I also own end-to-end ML/AI delivery beyond inference: retrieval-augmented generation work, real-time multilingual voice pipelines, and practical MLOps from data ingestion to deployment. I built a bilingual voice pipeline routing XTTS (GPU) and Kokoro (CPU sidecar) by detected language, resolving a silent-output failure traced to a swallowed generation-config error.

My approach is production-first: I built a semantic cache gateway to bypass GPU compute for repeated queries with prompt-injection defense and dual-tier rate limiting. Earlier, I’ve owned end-to-end ML pipelines on Databricks at Xorbix, improved RAG response quality by 24% through evaluation-driven tuning, and shipped customer-facing apps and pipelines from prototype through versioned rollout.

Experience

Work history, roles, and key accomplishments

Current

ML/AI Team Lead

Current

TekkDev

Aug 2025 - Present (11 months)

Built and deployed a production LLM platform on a single RTX 6000 (24 GB) and served Mistral 7B AWQ via vLLM with continuous batching on constrained hardware. Developed a real-time bilingual voice pipeline with XTTS and Kokoro and built a semantic cache gateway with prompt-injection defense and dual-tier rate limiting.

fastAPI Docker PostgreSQL VLLM Kokoro Prompt Injection Rate Limits

ML/AI Engineer Intern

Advangram

Apr 2025 - Jul 2025 (3 months)

Fine-tuned and deployed a text-to-SQL model to enable non-technical users to query internal data via natural language. Built and shipped forecasting and anomaly detection models to client-facing production apps using modular, version-controlled ML deployment pipelines.

Text To SQL Fine Tuning Forecasting Anomaly Detection Version Control

Data Scientist (ML/AI)

Xorbix Technologies Inc

Oct 2023 - Aug 2024 (10 months)

Owned end-to-end ML pipelines on Databricks for manufacturing and fintech clients, from data ingestion through production inference. Built an LLM evaluation framework to score RAG quality and tuned chunking strategies for improved response quality, and optimized FAISS GPU vector search for faster query performance.

Databricks ML Pipelines LLM Evaluation RAG Faiss Memory Management

Software Engineer

Siena College Actors' Database

Sep 2022 - May 2023 (8 months)

Built a full-stack web app using AWS and Django with query optimization and caching to improve response times. Optimized MySQL queries and indexing across core workflows and served as Scrum Master for a team of five using agile methodologies.

AWS Django Caching Query Optimization MySQL Indexing Scrum Agile Methodologies

Education

Degrees, certifications, and relevant coursework

Siena University

Bachelor of Science, Computer Science

Activities and societies: Minor in Data Science; Databricks Machine Learning Professional; GenAI Fundamentals certifications.

Earned a Bachelor of Science in Computer Science with a minor in Data Science. Completed Databricks Machine Learning Professional and GenAI Fundamentals certifications.