Abubakar Akram

Selected work

• 01

  AuditAhead

  Multi-agent pre-submission QA for Ontario pharmacy claims, producing a tamper-evident evidence packet an auditor can verify without trusting the host.

  In developmentPythonFastAPILangGraphAzure OpenAI
  →
• 02

  Mirlin

  A multi-tenant agentic RAG assistant that answers from each tenant's own documents and spreadsheets — with images, tables, follow-up memory and a sub-5-second target.

  Private.NET 8Bot Framework v4Semantic KernelAzure OpenAI
  →
• 03

  SDLC Agent Platform

  Five specialised agents drive a plain-language brief to shipped code, with a mandatory human gate at every stage.

  PrivateTypeScriptNode.jsLangGraphClaude Agent SDK
  →
• 04

  Volcanic Retail

  A two-sided B2B marketplace connecting brands with retail buyers — where I rebuilt the backend for speed and widened how the two sides find each other.

  LiveReactTypeScriptDjangoMySQL
  →

Experience

1. Sep 2025 — Present

   Current

   Senior AI Engineer · AIM Inc.

   Full-time · Lahore, Pakistan · Remote

   • Architected a compliance-grade multi-agent platform for a regulated industry — four specialised agents under supervisor orchestration, human-in-the-loop approval gates on high-risk actions, and a cryptographically signed, tamper-evident evidence trail so every agent decision can be reconstructed in an audit.
   • Shipped a multi-tenant agentic RAG platform letting business users query their own documents, tables and figures in natural language: hybrid retrieval on Azure AI Search, multi-index routing, and document-intelligence pipelines for layout, table and figure extraction. Every answer returns with its source citation.
   • Rebuilt agent orchestration on Semantic Kernel and AutoGen with tool calling, guardrails and deterministic fallbacks — raising task success 30% and eliminating the runaway failure loops that had been burning tokens on unrecoverable runs.

   LangGraphSemantic KernelAutoGenAzure OpenAIFastAPI.NET 8Azure AI SearchCosmos DB

2. Feb 2023 — Jul 2025

   Full-Stack AI Engineer · Calitech AI LLC

   Independent contractor · Remote, US

   • Fine-tuned LLMs with QLoRA and built LangChain RAG pipelines for Rad AI, a HIPAA-compliant clinical platform — cutting context-aware reporting time 25%.
   • Built end-to-end MLOps infrastructure with Kafka pipelines and Airflow-orchestrated workflows spanning ingestion, training, evaluation and deployment — cutting model latency 60% and deployment cycles 50%.
   • Architected AWS infrastructure (ECS, RDS, S3, SageMaker) supporting distributed GPU training and real-time prediction APIs, while cutting cloud costs 15%.

   QLoRALangChainFastAPIKafkaAirflowAWS SageMakerReactHIPAA

3. Jun 2020 — Jan 2023

   Full-Stack Developer · Digiance LLC

   Remote, US

   • Built a sales-outreach analytics dashboard for brands, surfacing actionable buyer insights that raised early-adopter pitch success rates 35%.
   • Engineered an automated testing and release pipeline with Jenkins and Docker, improving release confidence and reducing regression incidents.
   • Deployed scalable full-stack applications on Google Cloud (Compute Engine, Cloud SQL, GKE) with high availability and tuned resource utilisation.

   ReactTypeScriptDjangoMySQLJenkinsDockerGCP

4. Jan 2019 — Jun 2020

   Associate Software Engineer · Devsinc

   Onsite · Pakistan

   • Developed and refactored responsive UIs for a banking application, increasing performance and maintainability 40%.
   • Secured all user-facing routes of a financial application with OAuth 2.0, OIDC and JWT, hardening authentication and session integrity.

   ReactReduxMaterial UIOAuth 2.0OIDCJWT

BSc Computer Science · COMSATS University of Information Technology · 2015 — 2019