- Sat Dec 13, 2025 2:20 pm#11635
====================================================================
PREPARATION GUIDE FOR THE POSITION
Data Scientist / Associate Manager / Manager – AI Lab
BRAC Bank PLC – Dhaka (Deadline: 14 December 2025)
====================================================================
1. UNDERSTAND THE ROLE AND EXPECTATIONS
• Core focus: design, develop and fine‑tune Large Language Models (LLMs) for financial use‑cases such as data analysis, report generation and predictive modelling.
• Build Retrieval‑Augmented Generation (RAG) pipelines that connect LLMs with internal/external financial data, knowledge bases and vector databases.
• Create and maintain domain‑specific knowledge repositories (financial research, credit data, regulatory documents).
• Collaborate daily with analysts, quantitative researchers and business stakeholders to translate requirements into AI solutions.
• Deliver end‑to‑end predictive models and credit‑scorecards, overseeing project scope, methodology, documentation and deployment.
• Apply software‑engineering best practices (Git version control, unit testing, code reviews, CI/CD).
• Stay current with the latest advances in LLMs, RAG, and fintech innovations.
2. KEY KNOWLEDGE AREAS TO MASTER
A. Large Language Models & Prompt Engineering
– Architecture of Transformer‑based models (GPT, LLaMA, Falcon, etc.).
– Fine‑tuning techniques: LoRA, PEFT, full‑model fine‑tuning, adapter layers.
– Prompt design, chain‑of‑thought prompting, few‑shot prompting for financial narratives.
B. Retrieval‑Augmented Generation (RAG)
– Vector‑store concepts (embeddings, similarity search).
– Popular vector DBs: Pinecone, Weaviate, Qdrant, Milvus, FAISS, Elasticsearch with k‑NN.
– Indexing pipelines for tabular, PDF, HTML and image documents.
– Hybrid retrieval (dense + sparse) for regulatory compliance data.
C. Credit Risk Modelling & Predictive Analytics
– Logistic regression, decision trees, gradient boosting (XGBoost, LightGBM, CatBoost).
– Scorecard development (WOE, IV, monotonicity constraints).
– Model validation, KS‑test, ROC‑AUC, PD/LGD estimation.
– Explainability tools: SHAP, LIME, Counterfactuals.
D. Data Engineering for Finance
– Structured data handling (SQL, PostgreSQL, Snowflake, Oracle).
– Real‑time streaming (Kafka, Pulsar) for transaction feeds.
– Data cleaning, outlier detection, imputation for credit datasets.
E. Cloud & GPU Infrastructure
– Experience with AWS (SageMaker, EC2‑GPU, S3), Azure (ML, Blob), GCP (Vertex AI, Cloud Storage).
– Containerisation (Docker) and orchestration (Kubernetes) for scalable training/inference.
– Cost‑optimisation strategies (spot instances, mixed‑precision training).
F. Software Development Practices
– Git branching models (GitFlow, trunk‑based).
– Unit/Integration testing (pytest, unittest).
– CI/CD pipelines (GitHub Actions, GitLab CI, Azure DevOps).
G. Domain Knowledge – Banking & Financial Regulations
– Understanding of SME, corporate and retail banking products.
– Familiarity with Bangladesh Bank guidelines, Basel‑III, AML/KYC norms.
– Insight into credit underwriting workflows and risk appetite frameworks.
3. BUILD A TARGETED PORTFOLIO
• LLM Project – Fine‑tune an open‑source LLM on a curated set of financial reports (annual statements, credit policies) and demonstrate generation of a concise risk summary.
• RAG Demo – Create a pipeline that retrieves relevant sections from a corpus of Bangladeshi banking regulations and feeds them to an LLM for question answering. Use a vector DB and show latency/accuracy metrics.
• Credit Scorecard – Develop a full credit‑scoring model (WOE transformation → Logistic regression) on a public credit dataset (e.g., German Credit) and document the end‑to‑end workflow, validation, and deployment as a REST API.
• End‑to‑End Deployment – Containerise one of the above solutions, push to a cloud registry, and expose via a secure API gateway. Include monitoring (Prometheus, Grafana) and logging.
• Publish code on GitHub with clear READMEs, notebooks, and automated tests.
4. RESUME & APPLICATION TUNING
1. Header – Name, contact, LinkedIn & GitHub links, location (Dhaka).
2. Professional Summary – 2‑3 lines highlighting 1+ year experience with LLMs, RAG, credit risk modelling, and cloud‑GPU deployments. Mention purpose‑driven mindset and teamwork.
3. Core Competencies – List as bullet points: LLM fine‑tuning, Retrieval‑Augmented Generation, Credit Scorecard Development, Python, PyTorch/TensorFlow, SQL, Cloud (AWS/Azure), Vector Databases, Agile Project Management.
4. Experience – For each role:
– Action verb + specific achievement (e.g., “Designed a RAG system using Milvus and Open‑source LLaMA 13B, reducing query latency by 42 % for internal risk reports.”)
– Quantify impact (accuracy improvement, cost reduction, time saved).
5. Education – Degree, university, relevant coursework (Machine Learning, Statistics, Financial Engineering).
6. Certifications – Any recognised AI/ML or cloud credentials (e.g., AWS Certified Machine Learning – Specialty, Coursera “Generative AI with Large Language Models”, CFA Level 1).
7. Projects – Include the portfolio items with brief descriptions and links.
8. Additional Sections – Publications, talks, hackathon wins (especially fintech or AI‑focused).
5. INTERVIEW PREPARATION
Technical Screening
– Review fundamentals of Transformers, attention mechanisms, and LLM scaling laws.
– Practice coding LLM fine‑tuning scripts (PyTorch Lightning, Hugging Face Trainer).
– Implement a simple RAG flow from scratch (embedding generation → FAISS index → retrieval → prompt).
– Solve credit‑risk case studies: data preprocessing, WOE binning, model validation.
– Brush up on SQL queries, data‑pipeline design, and cloud‑service APIs (S3, SageMaker).
Behavioral / Leadership
– Prepare STAR stories that showcase:
* Collaboration with analysts to translate business needs into AI solutions.
* Managing a project deadline under tight constraints.
* Learning a new technology (e.g., vector DB) and delivering results.
– Align personal values with BRAC Bank’s purpose‑driven, inclusive and sustainability focus.
Domain Knowledge
– Read recent BRAC Bank annual reports, sustainability statements, and press releases.
– Study Bangladesh’s banking regulatory framework, especially sections on credit risk and SME financing.
Mock Sessions
– Pair‑program a small RAG prototype with a peer.
– Conduct a whiteboard explanation of a credit scorecard pipeline, emphasizing interpretability and governance.
6. CONTINUOUS LEARNING RESOURCES
*Books & Papers*
– “Attention Is All You Need” (Vaswani et al., 2017)
– “Retrieval‑Augmented Generation for Knowledge‑Intensive NLP Tasks” (Lewis et al., 2020)
– “Credit Scoring Using Machine Learning” (Hand & Henley, 2022)
*Online Courses*
– Coursera “Generative AI with Large Language Models” (DeepLearning.AI)
– Udemy “Building Retrieval‑Augmented Generation Applications”
– edX “Financial Risk Management in Emerging Markets”
*Technical Blogs & Repos*
– Hugging Face Transformers tutorials (fine‑tuning, LoRA)
– LangChain documentation for RAG pipelines
– VectorDB community examples (Milvus, Qdrant)
*Communities*
– AI‑Bangladesh Meetup groups (LinkedIn, Meetup)
– Kaggle “Credit Scoring” competitions for practice data
– GitHub “awesome‑llm” and “awesome‑retrieval‑augmented‑generation” lists
7. LOGISTICS & FINAL CHECKS
• Deadline is 14 Dec 2025 – submit the application well before the cutoff.
• Ensure all links (GitHub, portfolio, LinkedIn) are publicly accessible.
• Prepare a one‑page PDF cover letter that directly references the key responsibilities and highlights your most relevant experience.
• Verify the application email/URL: https://hotjobs.bdjobs.com/jobs/bracban ... ank883.htm
• Keep a copy of all submitted documents for follow‑up.
8. DAY‑OF‑INTERVIEW QUICK TIP LIST
– Dress business‑formal, maintain a friendly yet professional demeanor.
– Bring a printed copy of your resume and a QR code linking to your portfolio.
– Have a notepad ready for sketching architecture diagrams.
– Speak clearly about the impact of your past projects, using numbers.
– Show enthusiasm for BRAC Bank’s mission of financial inclusion and sustainability.
====================================================================
By following this structured preparation plan—strengthening core technical skills, showcasing relevant projects, tailoring your resume, and aligning with BRAC Bank’s values—you will present a compelling candidacy for the Data Scientist / Associate Manager / Manager – AI Lab role. Good luck!====================================================================
PREPARATION GUIDE FOR THE POSITION
Data Scientist / Associate Manager / Manager – AI Lab
BRAC Bank PLC – Dhaka (Deadline: 14 December 2025)
====================================================================
1. UNDERSTAND THE ROLE AND EXPECTATIONS
• Core focus: design, develop and fine‑tune Large Language Models (LLMs) for financial use‑cases such as data analysis, report generation and predictive modelling.
• Build Retrieval‑Augmented Generation (RAG) pipelines that connect LLMs with internal/external financial data, knowledge bases and vector databases.
• Create and maintain domain‑specific knowledge repositories (financial research, credit data, regulatory documents).
• Collaborate daily with analysts, quantitative researchers and business stakeholders to translate requirements into AI solutions.
• Deliver end‑to‑end predictive models and credit‑scorecards, overseeing project scope, methodology, documentation and deployment.
• Apply software‑engineering best practices (Git version control, unit testing, code reviews, CI/CD).
• Stay current with the latest advances in LLMs, RAG, and fintech innovations.
2. KEY KNOWLEDGE AREAS TO MASTER
A. Large Language Models & Prompt Engineering
– Architecture of Transformer‑based models (GPT, LLaMA, Falcon, etc.).
– Fine‑tuning techniques: LoRA, PEFT, full‑model fine‑tuning, adapter layers.
– Prompt design, chain‑of‑thought prompting, few‑shot prompting for financial narratives.
B. Retrieval‑Augmented Generation (RAG)
– Vector‑store concepts (embeddings, similarity search).
– Popular vector DBs: Pinecone, Weaviate, Qdrant, Milvus, FAISS, Elasticsearch with k‑NN.
– Indexing pipelines for tabular, PDF, HTML and image documents.
– Hybrid retrieval (dense + sparse) for regulatory compliance data.
C. Credit Risk Modelling & Predictive Analytics
– Logistic regression, decision trees, gradient boosting (XGBoost, LightGBM, CatBoost).
– Scorecard development (WOE, IV, monotonicity constraints).
– Model validation, KS‑test, ROC‑AUC, PD/LGD estimation.
– Explainability tools: SHAP, LIME, Counterfactuals.
D. Data Engineering for Finance
– Structured data handling (SQL, PostgreSQL, Snowflake, Oracle).
– Real‑time streaming (Kafka, Pulsar) for transaction feeds.
– Data cleaning, outlier detection, imputation for credit datasets.
E. Cloud & GPU Infrastructure
– Experience with AWS (SageMaker, EC2‑GPU, S3), Azure (ML, Blob), GCP (Vertex AI, Cloud Storage).
– Containerisation (Docker) and orchestration (Kubernetes) for scalable training/inference.
– Cost‑optimisation strategies (spot instances, mixed‑precision training).
F. Software Development Practices
– Git branching models (GitFlow, trunk‑based).
– Unit/Integration testing (pytest, unittest).
– CI/CD pipelines (GitHub Actions, GitLab CI, Azure DevOps).
G. Domain Knowledge – Banking & Financial Regulations
– Understanding of SME, corporate and retail banking products.
– Familiarity with Bangladesh Bank guidelines, Basel‑III, AML/KYC norms.
– Insight into credit underwriting workflows and risk appetite frameworks.
3. BUILD A TARGETED PORTFOLIO
• LLM Project – Fine‑tune an open‑source LLM on a curated set of financial reports (annual statements, credit policies) and demonstrate generation of a concise risk summary.
• RAG Demo – Create a pipeline that retrieves relevant sections from a corpus of Bangladeshi banking regulations and feeds them to an LLM for question answering. Use a vector DB and show latency/accuracy metrics.
• Credit Scorecard – Develop a full credit‑scoring model (WOE transformation → Logistic regression) on a public credit dataset (e.g., German Credit) and document the end‑to‑end workflow, validation, and deployment as a REST API.
• End‑to‑End Deployment – Containerise one of the above solutions, push to a cloud registry, and expose via a secure API gateway. Include monitoring (Prometheus, Grafana) and logging.
• Publish code on GitHub with clear READMEs, notebooks, and automated tests.
4. RESUME & APPLICATION TUNING
1. Header – Name, contact, LinkedIn & GitHub links, location (Dhaka).
2. Professional Summary – 2‑3 lines highlighting 1+ year experience with LLMs, RAG, credit risk modelling, and cloud‑GPU deployments. Mention purpose‑driven mindset and teamwork.
3. Core Competencies – List as bullet points: LLM fine‑tuning, Retrieval‑Augmented Generation, Credit Scorecard Development, Python, PyTorch/TensorFlow, SQL, Cloud (AWS/Azure), Vector Databases, Agile Project Management.
4. Experience – For each role:
– Action verb + specific achievement (e.g., “Designed a RAG system using Milvus and Open‑source LLaMA 13B, reducing query latency by 42 % for internal risk reports.”)
– Quantify impact (accuracy improvement, cost reduction, time saved).
5. Education – Degree, university, relevant coursework (Machine Learning, Statistics, Financial Engineering).
6. Certifications – Any recognised AI/ML or cloud credentials (e.g., AWS Certified Machine Learning – Specialty, Coursera “Generative AI with Large Language Models”, CFA Level 1).
7. Projects – Include the portfolio items with brief descriptions and links.
8. Additional Sections – Publications, talks, hackathon wins (especially fintech or AI‑focused).
5. INTERVIEW PREPARATION
Technical Screening
– Review fundamentals of Transformers, attention mechanisms, and LLM scaling laws.
– Practice coding LLM fine‑tuning scripts (PyTorch Lightning, Hugging Face Trainer).
– Implement a simple RAG flow from scratch (embedding generation → FAISS index → retrieval → prompt).
– Solve credit‑risk case studies: data preprocessing, WOE binning, model validation.
– Brush up on SQL queries, data‑pipeline design, and cloud‑service APIs (S3, SageMaker).
Behavioral / Leadership
– Prepare STAR stories that showcase:
* Collaboration with analysts to translate business needs into AI solutions.
* Managing a project deadline under tight constraints.
* Learning a new technology (e.g., vector DB) and delivering results.
– Align personal values with BRAC Bank’s purpose‑driven, inclusive and sustainability focus.
Domain Knowledge
– Read recent BRAC Bank annual reports, sustainability statements, and press releases.
– Study Bangladesh’s banking regulatory framework, especially sections on credit risk and SME financing.
Mock Sessions
– Pair‑program a small RAG prototype with a peer.
– Conduct a whiteboard explanation of a credit scorecard pipeline, emphasizing interpretability and governance.
6. CONTINUOUS LEARNING RESOURCES
*Books & Papers*
– “Attention Is All You Need” (Vaswani et al., 2017)
– “Retrieval‑Augmented Generation for Knowledge‑Intensive NLP Tasks” (Lewis et al., 2020)
– “Credit Scoring Using Machine Learning” (Hand & Henley, 2022)
*Online Courses*
– Coursera “Generative AI with Large Language Models” (DeepLearning.AI)
– Udemy “Building Retrieval‑Augmented Generation Applications”
– edX “Financial Risk Management in Emerging Markets”
*Technical Blogs & Repos*
– Hugging Face Transformers tutorials (fine‑tuning, LoRA)
– LangChain documentation for RAG pipelines
– VectorDB community examples (Milvus, Qdrant)
*Communities*
– AI‑Bangladesh Meetup groups (LinkedIn, Meetup)
– Kaggle “Credit Scoring” competitions for practice data
– GitHub “awesome‑llm” and “awesome‑retrieval‑augmented‑generation” lists
7. LOGISTICS & FINAL CHECKS
• Deadline is 14 Dec 2025 – submit the application well before the cutoff.
• Ensure all links (GitHub, portfolio, LinkedIn) are publicly accessible.
• Prepare a one‑page PDF cover letter that directly references the key responsibilities and highlights your most relevant experience.
• Verify the application email/URL: https://hotjobs.bdjobs.com/jobs/bracban ... ank883.htm
• Keep a copy of all submitted documents for follow‑up.
8. DAY‑OF‑INTERVIEW QUICK TIP LIST
– Dress business‑formal, maintain a friendly yet professional demeanor.
– Bring a printed copy of your resume and a QR code linking to your portfolio.
– Have a notepad ready for sketching architecture diagrams.
– Speak clearly about the impact of your past projects, using numbers.
– Show enthusiasm for BRAC Bank’s mission of financial inclusion and sustainability.
====================================================================
By following this structured preparation plan—strengthening core technical skills, showcasing relevant projects, tailoring your resume, and aligning with BRAC Bank’s values—you will present a compelling candidacy for the Data Scientist / Associate Manager / Manager – AI Lab role. Good luck!====================================================================
- By bdchakriDesk