Arshad Ahamed

Agentic AI Engineer · Applied AI Engineer · AI Solutions Engineer

+91 97317 01337· arshad@arshadify.online· Bangalore, India

portfolio.arshadify.online github.com/arshad98333 pangea.app/profile/arshad-ahamed linkedin.com/in/ArshaddAhamed

Available for remote roles worldwide · Relocation ready for GCC on site roles on a standard employment visa · Immediate start

Profile

Agentic AI engineer building governed, production grade AI on Azure. I design to a single rule: deterministic code owns money, limits and permissions; AI owns extraction, retrieval and explanation; humans approve high impact actions. Domain depth in GSO aligned food safety and trade compliance for the GCC, deployed sovereign on Azure UAE North. Recommended by contract clients 9 times.

Code owns money

Limits, permissions and payments stay deterministic and testable.

AI owns language

Extraction, retrieval, ranking and plain language explanation.

Humans own risk

High impact actions become an approval with evidence attached.

Results shipped

6% to 0.4%

Invoice error rate

after rebuilding billing on deterministic rules

6 days to 1

Month end close

finance reporting cycle time

11%

Lower stock loss

perishable inventory, forecasting led

INR 1.2L

Spoilage avoided monthly

valued at cost, sustained

47 hrs

Manual work removed monthly

purchasing, stock and reporting

10,000+

Hours of data supervised

agentic vision QA across 50+ volunteers

23%

Data quality lift

EEG collection sessions, Neuralace

Client recommendations

Top Earner status on Pangea

How I work

1. Business problem

Owner language, money and risk first

2. Clarify

Constraints, data, who approves what

3. Prototype

Thin slice in days, on real data

4. Solution

Deterministic rules, AI where it earns

5. Optimization

Evals, cost, latency, adoption

Experience

AI Engineer and ConsultantJan 2023 to Present

Arshadify, independent AI consultancy

Shipped an agentic operations stack for a regional food distribution client covering document extraction, reconciliation and reporting, cutting invoice errors from 6% to 0.4% and month end close from 6 days to 1 day.
Built demand and spoilage forecasting for perishable inventory, cutting stock loss by 11% and avoiding INR 1.2 lakh of spoilage per month; automated purchasing, stock and reporting workflows, removing 47 hours of manual work per month.

Freelance AI and Software Engineer

Contract clients via Pangea.app

Neuralace, EEG data collection platform: built an agentic vision system supervising data collection on local hardware. YOLO and MediaPipe track eye closure, blink rate, focus and distraction across 8 hour shifts, flagging non compliant sessions live and issuing session wise volunteer reports daily. Ran 50+ volunteers and 10,000+ hours of collected data, raising data quality 23%.
Pangea.app, talent marketplace: ran a security review of the production platform, found a critical vulnerability affecting user data, and delivered a disclosure with prioritised fixes the founding team shipped.
LeetDesign, system design platform: owned the platform end to end, from product structure and content architecture through build and release; authored a 22 problem system design curriculum across three difficulty tiers, spanning key value stores, rate limiters, session stores, payment ledgers and API gateways.
Client record: recommended 9 times, Top Earner status on Pangea, kudos for problem solving, communication, responsiveness, teamwork, reliability, innovation and independence.

Enterprise AI Selling, Applied

lyzr.ai cohort, Lyzr University, top performer

Top performer in lyzr.ai's enterprise AI selling cohort, then ran the motion: sourced 10,000 leads, scored them against a defined ICP with Trigify MCP, qualified 341 accounts at over 80% ICP match, and booked a pitch to an ecommerce marketplace of 600,000+ merchants.

Core skills

Agentic systems: Multi agent orchestration, LangGraph, LangChain, MCP tool servers, function and tool calling, planning, bounded autonomy, approval gates, audit trails.
Applied AI, LLMs and evaluation: RAG, hybrid and vector search (Azure AI Search), embeddings, reranking, grounding and citations, document intelligence, structured extraction, forecasting; golden datasets, eval set design, regression gates, tracing, token cost and latency budgets, drift monitoring.
Computer vision and edge AI: YOLO and MediaPipe pipelines, real time video inference on local hardware, blink, gaze and attention tracking, session telemetry and data quality control.
Programming: Python (async, typing, pytest), FastAPI, React, REST, MongoDB, SQL, Git; AI assisted development with Claude Code, Cursor and Copilot for spec driven build, refactoring and tests.
Guardrails, sovereign AI and cloud: Deterministic rule engines, least privilege, authorization layering, human in the loop, prompt injection defence, PII redaction, designed against EU AI Act, NIST AI RMF and UAE PDPL; sovereign deployment on Azure UAE North keeping data, embeddings, prompts and logs in region; Container Apps, Azure OpenAI, AI Search, Docker, CI/CD, OpenTelemetry, Auth0, API gateways, rate limiting, security review.

Projects

HAFIZAL GHIDHA · GCC cold chain and food compliance intelligencegidha.arshadify.online

Shipment, reefer and trade document workflows aligned to GSO food safety standards and UAE PDPL data handling. A deterministic rules engine issues accept, hold or reject; AI extracts documents and explains each decision in Arabic and English.

Grounded answers 93%Rule accuracy 98%Extraction 95%Human review 7%Cost per document INR 5Eval set 50 cases

Stack: Python, FastAPI, React, RAG, Azure AI Search, Azure Container Apps

Cold chain compliance, end to endAzure UAE NorthGSO aligned rulesHuman approval

Shipment and reefer sensors

Trade documents

AI extraction

Deterministic rules engine

Accept · Hold · Reject

Bilingual decision trace

MCPGATEWAY · Safe agentic AI for a telecom support deskmcpgateway.arshadify.online

Eight MCP tools behind eight ordered authorization checks that deny by default, on Azure Container Apps in UAE North. No refund path moves money autonomously: each becomes a supervisor approval with evidence attached.

Authorization precision 97%Test coverage 95%Tool calls traced 100%Autonomous money moved GBP 0Eval set 80 cases

Stack: MCP, FastAPI, MongoDB, Auth0, OpenTelemetry, Azure

Telecom support agent, end to endAzure Container AppsDeny by defaultNo money moved

MCP client

8 authorization stages

Guardrail pipeline

Tool gateway

Records and rules

Supervisor approval

Recognition and education

$101K in Azure founders credits won for a supply chain AI pitch; submission approved for Build with K2 Think V2, the MBZUAI hackathon in Abu Dhabi, with K2 Think API access.
Recommended 9 times by contract clients, with Top Earner status on Pangea.
Bachelor of Engineering, Computer Science, SJB Institute of Technology, Bangalore (November 2022).