Data Scientist

Valtech; Bengaluru; On-site; INR 1,500,000 - 2,000,000; Full time

Benefits offered by this job

Flexibility with remote work options

Career advancement opportunities

Access to training and industry experts

Job summary

Valtech in Bengaluru is seeking a Data Scientist passionate about pushing innovation boundaries in AI. You will develop models for fraud detection and work within Agile teams, collaborating closely with compliance and risk departments.

Applicants should have proven expertise in machine learning, especially in fraud settings, a strong foundation in data sciences, and experience managing complex models. This full-time role offers flexibility and opportunities for professional growth.

Qualifications

• 4+ years of experience in data science or related fields.
• Proven track record building fraud models.
• Experience in developing and operationalizing machine learning models.

Responsibilities

• Develop and maintain models for fraud detection.
• Collaborate with product teams on regulatory requirements.
• Expose model outputs via APIs for banking platforms.

Skills

Fraud model building

Graph analytics

Supervised learning

Computer Vision

NLP

Tools

PyTorch

TensorFlow

SQL

MLOps tools (MLflow, Kubeflow)

Job description

The opportunity

At Valtech, you’ll find an environment designed for continuous learning, meaningful impact, and professional growth. Whether you’re pioneering new digital solutions, challenging conventional thinking or building the next generation of customer experiences, your work will help transform industries.

As a Data Scientist, you are passionate about experience innovation and eager to push the boundaries of what’s possible. You bring 4+ years of experience, a growth mindset and a drive to make a lasting impact.

You will thrive in this role if you are:

• A curious problem solver who challenges the status quo
• A collaborator who values teamwork and knowledge‑sharing
• Excited by the intersection of technology, creativity and data
• Experienced in Agile methodologies and consulting (a plus)

Role responsibilities

• Develop, validate, and maintain supervised and unsupervised models for fraud detection, credit risk scoring, AML typology identification, and transaction anomaly detection.
• Build real‑time and near‑real‑time scoring pipelines integrating with banking event streams (Kafka, Pub/Sub) and decision engines.
• Perform deep exploratory analysis of transactional data, customer behavioural signals, merchant data, and graph‑based relationship networks to surface fraud patterns.
• Collaborate with compliance, risk, and product teams to translate regulatory requirements (RBI guidelines, PCI‑DSS, Basel III) into model design constraints.
• Construct and maintain feature stores covering entity‑level aggregations, velocity features, device/network signals, and geospatial behavioural attributes.
• Champion model interpretability using SHAP, LIME, and counterfactual explanations to satisfy audit and regulatory scrutiny.
• Design and fine‑tune LLMs (Gemini, GPT‑4o, Llama, Mistral) on proprietary banking corpora using PEFT and LoRA for tasks such as SAR narrative generation, dispute summarisation, and customer communication.
• Architect Retrieval‑Augmented Generation (RAG) systems grounded in internal knowledge bases — policy documents, fraud rulebooks, regulatory circulars — with vector stores (Pinecone, Milvus, Weaviate, ChromaDB).
• Apply Computer Vision and NLP to multimodal data pipelines (cheque images, KYC documents, audio call transcripts) for identity verification and fraud triage.
• Develop generative models (GANs, VAEs, Diffusion) for synthetic data augmentation to address class imbalance in fraud datasets while meeting data‑privacy obligations.
• Build prompt engineering frameworks using LangChain and LlamaIndex; implement chain‑of‑thought and agentic reasoning for complex investigative workflows.
• Write production‑ready Python code adhering to Valtech engineering standards — unit‑tested, type‑annotated, and reviewed.
• Operationalise models with MLOps tooling (MLflow, Kubeflow, Vertex AI Pipelines) covering versioning, A/B experimentation, drift monitoring, and automated retraining.
• Expose model outputs via FastAPI or Flask microservices integrated with banking middleware and case‑management platforms.
• Work within Agile delivery squads, participate in sprint planning, demo sessions, and client‑facing workshops.

Governance & Stakeholder Engagement

• Enforce Responsible AI principles — bias audits, fairness metrics, model cards, and NeMo Guardrails for deployed LLMs.
• Translate complex model behaviour into clear narratives for non‑technical stakeholders including compliance officers, fraud investigators, and C‑suite sponsors.

Must have qualifications

• Proven track record building fraud models (card‑not‑present, account takeover, synthetic identity, first‑party fraud, money‑mule networks).
• Experience with graph analytics (PyG, DGL, Neo4j) for network‑based fraud ring detection.
• Familiarity with banking data schemas: ISO 8583, SWIFT MT messages, core‑banking extracts, and bureau data (CIBIL/Experian).
• Exposure to regulatory frameworks: RBI Master Directions on Fraud, FATF AML/CFT guidelines, PCI‑DSS Level 1 environments.

Core AI / ML

• Expertise in supervised learning (XGBoost, LightGBM, neural networks) and unsupervised methods (isolation forest, autoencoders, DBSCAN) for anomaly detection.
• Strong foundations in Computer Vision and NLP; proven experience with multimodal pipelines combining images, text, and structured tabular data.
• Proficiency in PyTorch or TensorFlow for model development and custom training loops.

Generative AI Stack

• Hands‑on with Gemini, OpenAI GPT‑4, and open‑source LLMs (Llama 3, Mistral, Phi‑3).
• Model fine‑tuning using PEFT, LoRA, and QLoRA on domain‑specific corpora.
• RAG architecture design: chunking strategies, hybrid retrieval (BM25 + dense), re‑ranking, and query routing.
• Vector database proficiency: Pinecone, Milvus, Weaviate, or ChromaDB for semantic search and knowledge grounding.
• Advanced prompt engineering: chain‑of‑thought, few‑shot, structured output, and tool‑calling patterns using LangChain or LlamaIndex.
• Python (primary): pandas, NumPy, scikit‑learn, PySpark; clean, production‑ready, PEP‑8 compliant code with test coverage.
• Cloud: hands‑on experience with GCP (Vertex AI, BigQuery, Dataflow, Cloud Run), AWS (SageMaker, Redshift), or Azure (ML Studio, Synapse).
• SQL proficiency for complex analytical queries across relational and columnar stores (BigQuery, Snowflake, Redshift).

Nice to have qualifications

• MLOps tooling: MLflow, Kubeflow, Vertex AI Pipelines for end‑to‑end model lifecycle management.
• API Development: wrapping models in production REST APIs using FastAPI or Flask.
• Reinforcement Learning from Human Feedback (RLHF) and self‑supervised learning approaches.
• Experience with emerging GenAI architectures: multi‑agent systems, mixture‑of‑experts, speculative decoding.
• Databricks (Unity Catalog, Delta Live Tables, MLflow) for large‑scale feature engineering and model serving.
• Exposure to open‑banking APIs and real‑time payment rails (UPI, IMPS, RTGS) from a data perspective.

We do not require information such as age, gender, marital status, or a headshot in your application. We review all candidates based on skills, experience, and potential.

Commitment to reaching all kinds of people

We design experiences that work for all kinds of people – and that starts with our own teams. At Valtech, we’re intentional about building an inclusive culture where everyone feels supported to grow, thrive and achieve their goals. No matter your background, you belong here.

The benefits

This is a Full‑Time position based in Bengaluru.

Beyond a competitive compensation package, we offer:

• Flexibility, with remote and hybrid work options (country‑dependent)
• Career advancement, with international mobility and professional development programs
• Learning and development, with access to cutting‑edge tools, training and industry experts