Work

Five years across fintech, machine learning, and computer vision. The notable shipped work and the lessons each role left behind.

Zamp Finance

Zamp Finance

Backend Engineer
Full Time

Building reliable backend services for treasury and corporate finance — the kind of systems where the cost of being wrong is high enough to keep us honest. Working on infrastructure that scales horizontally without losing transactional correctness.

Stack
GoFastAPIGCPK8sNext.jsNew Relic
Highlights
  1. Owning core backend services that process treasury operations end-to-end — designing for both correctness and scale.
  2. Working across Go services and Python ML utilities, with a strong emphasis on idempotency, retries, and observability via New Relic.
  3. Contributing to the platform vision: making finance work at infinite scale, with reliability as a first-class concern.

Refyne India

Refyne India

SDE — 2 (Early Member)
Full TimeInternship → FTE

Joined as an early intern, promoted to SDE — 2 in 1.5 years. Spent three years building the infrastructure behind India's largest earned-wage-access platform. KYC, rewards, integrations, analytics — the unglamorous backbone that quietly held the fintech together. Led teams across B2C and Collections verticals while mentoring juniors.

Stack
NestJSFastAPIAWSK8sNext.jsElasticPostgresRedis
Highlights
  1. KYC engine — built from scratch. Verification time dropped from 3 minutes to 2 seconds. Conversion rates lifted from 20–25% to 70–75% per day, outperforming industry standards. Processed 500,000+ users in a year with zero reported issues.
  2. Rewards system — designed and shipped from scratch including scratch cards and virtual rewards. Organised the first Refyne festival, hitting a record DAU of 70,000 in 2023 and driving a 35% lift in user retention.
  3. Major client integrations — led integrations with Swiggy (₹102 cr disbursed to delivery partners) and Teleperformance. Designed Data Pipeline Frameworks processing 1M+ unstructured data points in under a minute.
  4. Refyne Analytical Platform — built an ETL pipeline converting NoSQL → SQL with a maximum lag of 2 minutes. Dashboard load times went from ~4 minutes to ~5 seconds.
  5. Leadership across B2C & Collections verticals — managed engineers, mentored juniors, owned roadmap and delivery.
  6. Promoted from Intern → SDE-1 → SDE-2 in 1.5 years on the back of consistent shipping and reliability ownership.
Press & Links

Omdena

Omdena

Machine Learning Engineer
InternshipRemote

Worked with Omdena's collaborative AI community on two challenges: a satellite-imagery model to identify the spread of illegal dumpsites globally, and a medical imaging model classifying pathologies from ultrasound scans.

Stack
PythonPyTorchAWSComputer VisionNLP
Highlights
  1. Dumpsite detection — trained CV models on multi-spectral satellite imagery to identify illegal waste accumulation across regions, contributing to environmental monitoring tooling.
  2. Medical ultrasound classification — built and fine-tuned models for pathology + mass identification from raw ultrasound scans.
  3. Collaborated remotely with ML engineers across 6+ time zones — a crash course in async engineering before it became the norm.

Precision Electronics

Precision Electronics

Computer Vision Intern
InternshipRemote

First professional CV work. Built a face-recognition attendance system designed to run on edge hardware — no cloud round-trip, no privacy concerns about uploading employee faces.

Stack
PythonOpenCVRaspberry Pi
Highlights
  1. Designed the full pipeline: face detection → embedding → matching against a local enrollment database, with retraining on new staff added.
  2. Deployed and validated on a Raspberry Pi 4 in a real office setting — proved the system held up under variable lighting and angles.
  3. Where I learned that ML in production is 10% model and 90% glue code.
— end of record —
For the full résumé, a chat, or anything else: divymohanrai@gmail.com