Machine Learning Engineer — Bidding & Prediction

Who we are

AdZoic is an ad-buying platform. Brands and agencies tell us their goal and budget, and our AI buys the right ads for them, in real time — on streaming TV, online video, websites, apps, audio and digital billboards. We're a small team in Dhaka, and we build and run all of it ourselves.

Why this role matters

Our AI is our product. When an advertiser says “get me customers for this budget,” it's our models that decide which impressions to buy and what to pay — millions of times a day, in real time. You'll build that intelligence — models that predict clicks and conversions, price each bid, pace budgets and catch fake traffic — and take them all the way into the bidder, where they must answer in a few milliseconds.

What you'll do

• Build and ship the models at the heart of our bidding: click prediction (CTR), conversion prediction (CVR), bid pricing and budget pacing.
• Build fraud-detection models that spot fake clicks and bot traffic before we bid.
• Turn billions of raw ad events into clean features and training data — build the pipelines, not just the models.
• Serve your models inside the bidder at very low latency — you'll learn to trade accuracy for speed and still win the auction.
• Run honest A/B tests: ship a model, measure its real impact on revenue and campaign results, keep it or kill it on the numbers.
• Watch your models in production — track drift, retrain on schedule and catch it fast when predictions go bad.
• Work closely with the bidder engineers (Go) on serving, and with the data team on event pipelines.

What you'll bring

• 3+ years in ML engineering or data science, with models you personally put into production — not just trained.
• Strong Python and hands-on experience with standard ML frameworks (scikit-learn, XGBoost/LightGBM, PyTorch or similar).
• Solid statistics and experimentation skills — you know when a result is real and when it's noise.
• Experience working with large datasets: strong SQL, plus Spark or similar for heavy processing.
• An understanding of what low latency demands from a model — size, features and serving all have a budget.
• Ownership — you ship it, you run it, you fix it.

Nice to have

• Experience with real-time feature serving or streaming pipelines (Kafka, ClickHouse, Redis or similar).
• Go experience — enough to work alongside the bidder team on the serving path.
• AdTech or OpenRTB experience — not required; strong engineering matters more.

Schedule & location

• 11:00 AM – 7:00 PM (Bangladesh time), Sunday to Thursday.
• Hybrid — 3 days in the office, 2 days from home.
• House 42, Road 08, Block D, Niketan, Gulshan, Dhaka 1212.

What you'll get

• Salary — negotiable.
• Festival bonus (twice a year), plus performance and referral bonuses.
• Medical coverage for you and your immediate family (in process).
• Learning budget for courses, certificates and conferences.
• Quarterly team outings; annual, casual and sick leave, plus maternity and paternity leave.