Senior AI Engineer

Role Overview

Own the end-to-end build of an AI chatbot for financial insights, including continuous data ingestion, autoscalable serving, AI-based stock/company summarization, and user profiling integrated with analytics platforms.

What Youll Do

1) Chatbot based on data pooling / retrieval

• Design a

  RAG pipeline

  over pooled internal + external sources (filings, news, fundamentals, broker notes).
• Implement

  data chunking

  ,

  hybrid search

  (BM25 + vector),

  cross-encoder re-ranking

  ,

  query rewriting

  , and

  grounded citations

  .
• Stand up

  vector stores

  (Pinecone/Milvus/FAISS) with

  HNSW/IVF

  indexes; manage embedding updates and TTL.
• Build low-latency

  chat APIs

  (FastAPI/Node) with session memory, safety filters, and audit logs.

2) Continuous / frequent data ingestion

• Build

  streaming + batch

  ingestion for sources like NSE/BSE feeds, EDGAR/SEBI filings, news/RSS, fundamentals APIs.
• Use

  Kafka/Kinesis + Debezium

  for CDC,

  Airflow/Prefect

  for orchestration,

  Great Expectations

  for data quality.
• Store in

  S3/GCS + Parquet

  , Lakehouse (

  Delta/Iceberg

  ), with

  Glue/BigQuery

  catalogs and

  Schema Registry

  .
• Implement

  idempotent

  jobs, late-data handling, backfills, and SLAs/alerts.

3) Auto-scalable module

• Deploy on

  Kubernetes

  (EKS/GKE/AKS) using

  HPA/KEDA

  (scale on QPS, Kafka lag, GPU metrics).
• Containerize via

  Docker

4) AI-based summarization of stocks & parent companies

• Implement

  map-reduce / refine

  summarization with

  fact-checking

  , citation grounding, and

  temporal aggregation

  .
• Mix general + finance LLMs (

  GPT-4.1/4o-mini, Llama-3.1, Mistral-Large

  ) via adapters; use

  Ray Serve/BentoML

  for model serving.
• Factuality evals with

  QAFactEval/RAGAS

  , prompt/version management, and offline eval harnesses.

5) User profiling via analytics

• Integrate

  GA4

  (BigQuery export),

  Segment

  ,

  Mixpanel

  for events and cohorts; define

  PII-safe

  identity
• Build features for

  RFM

  , portfolio affinity, risk appetite,

  embedding-based personalization

  , and

  bandit-driven

• Respect

  consent/opt-out

  , implement

  RBAC

  , secrets via

  KMS/HashiCorp Vault

  , and data retention policies.

Must-Have Skills

• Python/TypeScript

  ,

  FastAPI(alternative framework)

  ,

  SQL

  ; strong data structures & systems

• RAG

  patterns,

  vector databases

  (Pinecone/Milvus/FAISS),

  embedding models

  (E5/Instructor/OpenAI).

• LLM prompt engineering

  ,

  summarization strategies

  (map-reduce/refine),

  re-ranking

  (cross-encoder,

• Observability

  (Prometheus, Grafana, OpenTelemetry) and

  monitoring/evals

  for LLM

• Security & compliance

  basics (PII handling, secrets, IAM/RBAC, audit logs).

Good to Have

• Ray Serve/BentoML/Triton

  ,

  Torch/TensorRT

  , GPU scheduling on k8s.

• Feature stores

  (Feast),

  A/B testing

  ,

  bandits

  , offline/online
• Finance domain: fundamentals, corporate actions, tick/quote
• Experience with

  BigQuery/Redshift

  ,

  dbt

  ,

  Looker

• Load testing (

  Locust/k6

  ), chaos/resilience testing.