Skip to content

extraction run · cv → record

Calm hands on hard problems.

I'm Ahamed, a forward-deployed AI engineer from Colombo. I build AI-native platforms and transform traditional software with AI — turning the documents people re-key by hand into structured, useful work.

So here is my CV, read the way I'd read yours.

See the proof Download CV

Ahamed Musharaf

Forward-deployed AI engineer · Colombo, Sri Lanka

Experience

AI Software Engineer — Sansa Technologies · Dec 2025 – Present

Specialty-insurance intake cut from ~45 min to under 5 min per submission

~70% of fields auto-populated · 6+ ACORD forms generated in seconds

Software Engineer (AI & Full-stack) — SimpleBooks · Sinhala / Tamil OCR

Machine Learning Engineer — Absol X · AI sales agent + ERP

Education

MSc Data Science & AI — University of Moratuwa · 2026

BSc Electrical Engineering — NUST · Allama Iqbal HEC full scholarship

Tools

Claude · OpenAI · Gemini · Claude Code · MCP · RAG · Temporal · LLM evals

Ahamed Musharaf

record · ahamed_musharaf

Ahamed Musharaf

role
Forward-deployed AI engineer
now
AI Software Engineer @ Sansa · MSc @ Moratuwa
impact
45m → <5minsurance intake, per submission
builds
AI-native platforms · document AI · computer vision
models
Claude · OpenAI · Gemini
based
Colombo, Sri Lanka
open_to
Forward-deployed & applied AI roles

7 fields · every value traceable to the CV

raw_input0/7

01rolesource · CV · current work

New AI-native platforms, and old systems taught new tricks.

New platforms or existing systems, the aim is the same: less repetitive work, and software that makes people's days easier. Multi-agent systems are one part of the toolkit, not the whole story.

  1. 01

    AI-native platforms & transformation

    New platforms built around AI, and traditional systems reworked to automate the busywork. Multi-agent systems, retrieval and tool use are ways to connect intelligence to action—not the limits of what I build.

    • Claude
    • OpenAI
    • Gemini
    • Multi-agent
    • MCP
    • RAG
    • LLM evals
    • Claude Code
    • Temporal
  2. 02

    Workflow automation & document AI

    Turning manual, document-heavy work into typed pipelines — including multilingual OCR that reads Sinhala and Tamil.

    • Extraction
    • Multilingual OCR
    • Typed APIs
    • Anvil
    • Postmark
    • HazardHub
  3. 03

    Computer vision & ML

    Real-time detection, depth estimation and image-to-image models — from research prototype to something that runs at speed.

    • YOLO
    • Depth
    • pix2pix
    • PyTorch
The full toolbox 43 tools, grouped by the work they do

AI & agentic development

  • Claude (Opus · Sonnet · Haiku)
  • Claude Code
  • Repository skills
  • Multi-agent orchestration
  • MCP
  • RAG
  • Prompt caching
  • Context management
  • Vercel AI SDK
  • OpenAI
  • Gemini
  • Artificial Intelligence (AI)
  • Generative AI
  • Prompt Engineering

Applications & workflows

  • TypeScript
  • React
  • FastAPI
  • Express
  • Temporal
  • oRPC
  • Zod
  • Clerk

Data & connected services

  • Postgres · Drizzle
  • AWS S3
  • Cloudflare
  • Postmark
  • Anvil
  • HazardHub
  • Amazon Web Services (AWS)

Evaluation, testing & observability

  • LLM evals
  • Grounding checks
  • Langfuse
  • OpenTelemetry
  • Vitest
  • PGlite

Document intelligence & vision

  • Document AI
  • Multilingual OCR
  • Computer vision
  • YOLO
  • Depth estimation
  • PyTorch
  • Natural Language Processing (NLP)
  • Machine Learning (ML)

02proofsource · CV · 4 roles

Before, and after.

Four stories, one pattern: find the unglamorous bottleneck, and engineer it out of the workday. Every number here is reported in my CV, with scope and limits in each case study.

Sansa Technologies · AI Software Engineer · Dec 2025 — Present

Less paperwork. A faster placement.

Specialty insurance submissions arrive as a pile of documents. Brokers, underwriters and carriers were re-keying the same data by hand — ~45 minutes per submission, across half a dozen forms.

Intake time per submission
45m → <5m
Fields auto-populated
~70%
Forms generated in seconds
6+
Inside the project The platform around it ↗

Before · manual intake

  1. 01Incoming documents + property & hazard data
  2. 02The same details typed again, field by field
  3. 03ACORD forms and carrier supplementals filled one by one

~45 minper submission, re-keyed by hand

After · the delivered workflow

  1. 01Documents + OCR, enriched with HazardHub
  2. 02AI extraction → ~70% of fields auto-populated
  3. 03Anvil: 6+ ACORD forms + e-signature in seconds
  4. 04Carrier appetite matching to eligible markets

<5 minper submission, as reported in my CV

Drag to compare. Results reported in my CV · a simplified sketch, not a production trace.

SimpleBooks (Pvt) Ltd. · Apr 2025 — Mar 2026

An AI tax consultant that reads your paperwork

Before

Tax prep and company registration meant slow, manual data entry from payslips, T-10 forms and national IDs — with no room for the two other national languages people actually use.

After
  1. NIC image · Sinhala / Tamil
  2. Multilingual OCR
  3. Structured director details
  4. Custom API
  5. Registration frontend
Inside the project

Absol X · Feb 2025 — Mar 2025

From a conversation to an order.

Before

A car dealership needed an AI sales agent for onboarding, negotiation, test-drive appointments and order placement, alongside an ERP for its operations.

After
  1. Customer onboarding
  2. Price negotiation
  3. Test-drive appointment
  4. Order placement
  5. ERP / customer operations
Inside the project

Research & freelance · 2020 — 2025

Teaching machines to see

Before

Before the agents, there were pixels: surveillance, robotics, manufacturing and e-commerce problems that needed models fast and accurate enough to actually deploy.

After
  1. Camera / image input
  2. Model inference
  3. Detection / depth / measurement
  4. Application integration

20 FPS real-time detection+20% measurement precision−25% depth-model overhead

Inside the project

03methodsource · how I work · team practice

Forward-deployed means staying until it works.

I don't hand off a model and walk away. I sit with the people whose work the software is meant to change — and I stay until it does.

  1. 01

    Calm under pressure

    The system is on fire and the client is watching — that's exactly when a steady hand matters most.

  2. 02

    End-to-end ownership

    From the messy first prototype to the last mile of a deployed product. I own the whole arc.

  3. 03

    Humans in the loop

    The best AI knows when to ask a person. I design for confidence, review and trust — not blind automation.

In practice, at Sansa

  1. Give the agent a working brief.

    Repository skills for Claude Code organize domain context, contract rules, coding patterns and checks into task-specific guidance. Useful context is selected for the job, not treated as an unlimited prompt.

  2. Keep answers attached to evidence.

    The documented intake loop connects tool use to source material and citations, with a path for questions that need a person. A plausible answer and a supported answer are different things.

  3. Test the answer. Then question the test.

    The evaluation guidance puts deterministic checks before grounding checks and semantic assessment. Expectations must match what the agent was actually given—not what the test author meant to send.

Practices from the team platform's repository guidance, not solo claims. The platform context ↗

From the notebook · drafts

The playground · fictional · ~3 min

Take the pager for a migration under pressure.

Diagnose a legacy monolith, deploy the new pipeline, contain a live incident. Illustrative, no customer data.

Enter the engine room

04originsource · CV · education

Colombo → Islamabad → Colombo.

A full merit scholarship took me to NUST for four years of electrical engineering. The computer-vision and ML work started there, and came home with me.

Route from Colombo to Islamabad in 2020 for university, and back to Colombo from 2024.2020 · on a full merit scholarship2024 · home, with an engineering degreeColombo6.93° N · 79.86° EIslamabad33.68° N · 73.05° E

  1. 2019 Colombo

    Royal College, Colombo 07

    Grew up curious in Colombo — and I still call it home.

  2. 2020 Islamabad

    Allama Iqbal HEC full scholarship

    A full merit scholarship took me from Colombo to Islamabad.

  3. 2024 Islamabad

    BSc, Electrical Engineering · NUST

    Four years far from home — where the CV and ML work really started.

  4. 2025 Colombo

    Forward-deployed, three times over

    Sansa · SimpleBooks · Absol X — real clients and real deadlines.

  5. 2026 Colombo

    MSc, Data Science & AI · Moratuwa

    Going deeper on the theory behind the systems I build.

Credentials & training 6 items · AWS Academy badges are training, not exams

Training badge

AWS Academy Graduate - Generative AI Foundations

Amazon Web Services Training and Certification

Issued

View on Credly ↗

Training badge

AWS Academy Graduate - Machine Learning for Natural Language Processing

Amazon Web Services Training and Certification

Issued

View on Credly ↗

Training badge

AWS Academy Graduate - Machine Learning Foundations

Amazon Web Services Training and Certification

Issued

View on Credly ↗

MATLAB Certification

MathWorks

AWS Academy badges recognize completed training. If an embedded badge doesn't load, use its link to view the credential directly on Credly.

05personsource · off the clock

Serious about the work. Easy about everything else.

A little altitude. A lot of peace.I reset at altitude and on slow trains. New places, thin air, and long views are how I come back with clearer thinking.
Always curious about the next stop.New places. Different perspectives. Slow trains and long views.
MatchdayFootball is my off-switch. Nothing quite clears a hard problem like a full stadium and a game that refuses to be predictable.
A notebook that never closesCuriosity is the job. I keep a running notebook of ideas, sketches and half-built things — most of what I ship started as a scribble.

Selected imagery that reflects my interests, not photographs from my own trips.

06asksource · the observatory

Still curious? Ask the sky.

I charted my life as a night sky and taught an AI to navigate it. Ask a question and the right constellation lights up, or click a star and read it yourself.

Open the Observatory “Why should I hire you?”“What do you actually do all day?”

07outputsource · you

Let's build something that survives reality.

Recruiter, founder, or fellow builder — if you have a hard problem or a team worth joining, I'd love to hear about it.

Email me directlymtamusharaf@gmail.com
Connect on LinkedInExport this record · CV

Based in Colombo · happy to talk

Your details are used to respond to your enquiry.