Role: Senior Machine Learning Engineer

Experience: 6-8 Years

Location: Bangalore - Hybrid

Models in production. Not notebooks.

Candidate Brief

Who We're Looking For

We're not hiring a coder. We're hiring an engineer. We want someone who genuinely loves solving problems, thinks beyond the blocker in front of them, and reasons about the whole system rather than just the file they happen to be editing.

If your instinct when hitting a wall is to find a way around, over, or through it and then tell the team what you decided and why - keep reading.

What You'll Do

You'll own our propensity modelling and wider machine learning initiatives end to end - framing the problem, building the features, training and validating the model, getting it into production, and keeping it healthy once real decisions depend on it.

This isn't a research seat. The value of a model here is what it changes in the product and the business, so you'll care as much about the serving path, monitoring, and retraining loop as you do about offline model lift.

You'll make architectural calls on features, training, deployment, and monitoring - and ship them. You'll also work directly with product, data, and business stakeholders, communicating clearly, disagreeing well,

explaining models to non-technical audiences, and understanding the domain deeply.

What We Expect You to Be Robust At

Required

ML in production - Mandatory: You have taken models all the way into production and lived with them afterwards - serving, monitoring, retraining, and the pager. Notebooks, Kaggle placings, and coursework don't count on their own. You can talk about a model that degraded in the wild and what you did about it.

Propensity & customer modelling: Hands-on with propensity, churn, conversion, or similar customer outcome models. You know how to frame the target, handle heavy class imbalance, choose metrics that reflect the business decision rather than just AUC, and calibrate probabilities so a score means something.

Uplift and causal approaches are a real plus.

Feature engineering & data discipline: Feature stores, point-in-time correctness, and a healthy paranoia about target leakage and train/serve skew. Strong SQL and a warehouse-first instinct for where features should actually live.

Validation you can defend: Honest experimental design - proper splits, backtesting on time-ordered data, baselines, and knowing when a result is too good to be true. You measure impact in production with

A/B tests or holdouts, not just offline metrics.

MLOps & lifecycle: Reproducible training pipelines, experiment tracking, model registry and versioning,

CI/CD for models,



and automated retraining. You instrument for drift - data drift, concept drift, and performance decay - and know what to do when the alarm goes off.

Databricks - Preferred: Hands-on experience with Databricks, Spark, MLflow, Unity Catalog, Feature

Store, and Model Serving is a real advantage. Equivalent platforms are acceptable if you can map across quickly.

Compliance, governance & explainability: Model documentation and lineage, reproducibility, audit trails,

approval workflows, explaining why a model made a decision, testing for bias across groups, and designing with model risk and fair-treatment obligations in mind.

Security in a multi-tenant SaaS: Tenant isolation is a first-class concern - no data crossing customer boundaries through features, training sets, caches, or logs. PII handling and access control are part of how you design.

Strong engineering fundamentals: Production-quality Python and confident SQL. You write maintainable,

tested code and treat pipelines as software rather than scripts.

Systems thinking: You see how the pieces connect and where they'll break under load, over time, or at scale.

Production-grade, at pace: You enjoy building things that hold up in production and think in hours and days, not weeks and months. Speed and quality aren't a trade-off you accept - you find the version that's both.

Non-Negotiable: LLM Usage

You use LLMs to multiply your own output. We want someone already deep in using LLMs to unblock themselves and move faster - driving an ecosystem with LLMs, wiring them into workflow, tooling, and the way the team ships.

- Building at LLM speed - using models to prototype, refactor, and explore designs, with judgement to know when output is good enough to ship.
- Coding agents and dev tooling - real fluency with the current generation of AI development tools, and a view on where they help and where they get in the way.
- Staying current - tracking what has changed, trying things, and bringing the team along.

If you've only heard of these tools, or have dabbled but never worked with them in depth, you are not a fit. We need someone who has leaned on them enough to know their strengths, failure modes, and how to get real leverage out of them.

Talk Us Through Your Work





Demo is a bonus.

We care far more about depth than polish. We want to sit with you and go deep on something real you've worked on - how you thought about it, not how well you can pitch it. A demo is welcome but not expected.

- What you built and why - the problem, options considered, and why you chose your approach.
- What went wrong - where it failed, what surprised you, and what you'd do differently now.
- The numbers behind it - how you measured performance, what it did in production versus offline,

and how you improved it.
- The trade-offs you made - where you deliberately chose 'good enough,' and where you refused to compromise.

Nice to Have A plus

Fintech or financial-services experience: A huge plus. Knowing how banks and financial products think about customers, risk, and regulation means you'll frame better problems from day one. Not a hard requirement.

Marketing, CRM, or campaign analytics: Audience selection, targeting, and measuring incremental impact rather than vanity lift.

Streaming or near-real-time inference: Scoring in the moment, not just in nightly batches.

Vibe coding experience: You've built real things this way and know when it works and when it doesn't.

Strongly Preferred

Startup or small product company background.

We want startup thinking: ownership, urgency, comfort with ambiguity, and bias toward action.

Experience at larger service companies or slower-moving big banks is less preferred - not because of the name on the resume, but because we want people who haven't learned to sit behind bureaucracy and blockers.

What a Strong Fit Looks Like

- Has propensity or customer-modelling depth - can talk about targets, leakage, calibration, and measured business impact, not just model families.
- Has shipped models to production - trained, deployed, monitored, and retrained, with an end-to-end story they can walk through.
- Can go deep on real work - can explain a model shipped in concrete detail: target, pitfalls, and numbers.
- Ships production-grade and fast - thinks in hours and days and can point to things built that held up in production.
- Comes from a startup or small product company - bias to action over process. Fintech experience is a huge plus; profiles from large service companies or slower-moving banks need a clear signal they move fast and make decisions.

About the Company

Pace Wisdom Solutions has been working with Fortune 500 Enterprises and growth-stage startups/SMEs since 2012. We are a deep-tech Product engineering and consulting firm with offices in San Francisco, Bengaluru, and Singapore.

📌 Senior ML Engineer (Bengaluru)
🏢 Pace Wisdom Solutions
📍 Bengaluru