Section 2 of 7
Industry experience
Building trustworthy AI systems for finance workflows — grounded generation, evaluation, and monitoring.
Selected roles — focused on high-stakes finance contexts and measurable reliability.
Industry 01
AI/ML Engineer · Savantiq Jan 2026 – Present
Building the AI core of an investment intelligence platform (high-level public summary only).
Focus
- Unified fragmented LLM usage into a modular RAG platform (Python/Django) powering chat, memo generation, AI summaries, and research pages — with query optimisation, claim-based retrieval, and grounded answer synthesis with inline citation.
- Designed a claim-first retrieval architecture to reduce context contamination, integrating AWS Bedrock via a multi-provider AI client (Azure OpenAI + Bedrock) for more consistent answers.
- Migrated a legacy chat entrypoint into a SmartRouter → ChatPipeline architecture with a test and rollout plan, simplifying production operations.
- Built an evaluation + monitoring loop capturing router decisions, retrieved context, and user feedback to support proactive quality control.
Industry 02
Data Science Consultant · LSEG May 2022 – Jul 2023
Real-time market monitoring at infrastructure scale.
Outcomes
- Reduced critical incident detection time by 80% by leading the Operational Readiness Centre project: automated, real-time market health monitoring pipelines (AWS S3, Glue, SageMaker, ElasticSearch, Snowflake, ServiceNow).
- Built end-to-end data pipelines across 5+ source systems, delivering operational insights via PowerBI and Kibana to Capital Markets stakeholders.
- Prototyped a predictive delay-detection model (AWS SageMaker), extending the system from reactive monitoring to proactive incident prevention.
- Built an LSTM-based email classifier (85% accuracy) reducing mean client response time by 25%.
Industry 03
Data Science Intern · Stellar Fusion Aug 2023 – Oct 2023
Financial metrics analysis from SEC filings.
Delivered
- Built a Python/SQL pipeline to ingest, clean, and structure financial metrics from SEC filings and MongoDB company data across 500+ company records.
- Automated cross-sector metric comparisons using K-means clustering and TF-IDF NLP — presented as a proof of concept to senior stakeholders.
Core themes
◎
Grounded by default
Claim-first retrieval + citations when answers drive decisions.
▦
Measured reliability
Evaluation harnesses and monitoring loops, not vibes.
◇
Workflow-first delivery
Design systems around users, decisions, and failure modes.
