Available for new opportunities

I engineer Agentic AI and build robust Backend Systems.

Formal Graduate in Artificial Intelligence. Specializing in decoupled ML pipelines, LoRA fine-tuning, ReAct agents, and high-throughput server architecture.

View Work

Impact & Experience

2026 GLOBAL

1st Place Global - AutoScientist Challenge

Adaption Labs • Finance Category

Secured first place globally by architecting advanced predictive modeling and quantitative analysis pipelines. Outperformed an international pool of developers by focusing on optimization and highly tuned model reasoning rather than standard API abstraction.

AI Intern

SMIT (Saylani Mass I.T)

May 11 - Aug 11, 2026

Engineered decoupled transfer learning pipelines.
Optimized real-time vision inference (CPU/MobileNetV3).
Architected async FastAPI backend foundations.

Adaption Labs Ambassador

Selected for the inaugural cohort. Leading localized builder ecosystems and steering developers toward adaptive, real-time AI systems and open-source model optimization. Driving technical workshops on PyTorch and Agentic architecture.

Leadership DevRel Hackathons Community Building

Featured Engineering

Autonomous Agent

BidPilot

An advanced, autonomous bidding and analytical engine. Driven by a LangChain ReAct agent core, integrated with a high-performance FastAPI backend and ChromaDB for semantic memory storage, ensuring context-aware, real-time decision making.

FastAPI LangChain ReAct Agents ChromaDB

Source Code

LoRA Fine-Tuning

MedScout-17B & CodeIntel-v1

Domain-specific precision fine-tunes published to Hugging Face. MedScout-17B adapts Llama-4 for complex clinical visual QA. CodeIntel-v1 leverages Gemma-3 to optimize specific Python reasoning capabilities.

PyTorch LoRA / PEFT Hugging Face

Source Code

Air-Gapped Privacy

Local Docker RAG Assistant

A zero-hallucination QA system bounded strictly to verified Docker documentation. Built natively using LlamaIndex, ChromaDB, and Ollama to ensure complete data privacy—executing entirely on-device with zero cloud API dependencies. CPU-optimized chunking for fast inference.

LlamaIndex Ollama Streamlit RAG

View Repository

docker_rag.py

import chromadb

from llama_index.llms.ollama import Ollama

# Init local LLM

llm = Ollama(model="llama3.2:1b", request_timeout=300.0)

# Query embedded knowledge base

response = query_engine.query("How to optimize multi-stage builds?")

> Synthesizing answer based on 340 doc chunks...

Let's build something.

Open to roles in AI Engineering, Backend Architecture, and freelance opportunities.

dogarasad277@gmail.com

github.com/Asadullah-Dogar linkedin.com/in/asad-ullah-dogar