Anna University
Education, College of Engineering, Chennai, 2017 – 2021. B.E., Electronics & Communication.
Four years of signals and systems: how to pull a clean signal out of noise. Turns out that’s the whole job.
What I did
- Studied signals and systems, communication and electronics engineering.
- Built the maths habit: model the noisy real world before trusting it.
What I learned
- How to separate signal from noise, which turned out to be most of data work.
- To trust the maths over a hunch when the data is noisy.
LTIMindtree
Experience, Chennai, 2021 – 2022. Software Engineer.
An insurance client’s data arrived from everywhere. I built the pipelines that brought it together.
75% more efficient data flow; 30% less redundant data; 20+ client dashboards.
What I did
- Designed and optimized ETL pipelines in Informatica, SSIS and Azure Data Factory.
- Deployed and ran production ETL through CI/CD and Git, monitoring jobs and fixing failures.
- Built SSAS OLAP cubes and dimensional models, and 20+ Power BI and Tableau dashboards.
What I learned
- Production data breaks at night. Monitoring and clear fixes matter as much as the pipeline.
- Model data around the questions people ask, not the systems it came from.
- Shipping through CI/CD and Git makes data work repeatable.
Barclays
Experience, Chennai, 2022 – 2024. Data Analyst, Mortgage Product Owner.
I found the mortgage customers most likely to say yes, and automated the reports nobody wanted to build.
+40% campaign response; −70% reporting turnaround; +15% user retention.
What I did
- Product Owner for the mortgage portfolio, aligning data scientists and campaign managers with business goals.
- Segmented customers on transaction and engagement data; wasted marketing spend fell 5%.
- Built and ran 50+ campaigns on Teradata and CRM platforms, with Tableau and Excel dashboards.
- Replaced manual reporting with automated SQL and Python pipelines.
What I learned
- A segment only matters once the campaign team trusts it, so I learned to explain models in business terms.
- As Product Owner: get data scientists and marketers on one goal before building anything.
- Automating the dull reports frees people for the real analysis.
Northeastern
Education, Boston, 2024 – 2026. M.S., Data Analytics Engineering.
In 2024 I moved from Chennai to Boston for a master’s in data analytics engineering.
4.0 GPA, out of 4.0; Dec 2026 graduating.
What I did
- Coursework in Data Management, Computation & Visualization, Data Mining, MLOps, Deep Learning and Applied GenAI.
- Built MOMENTO for a Google-sponsored MLOps Expo and THE EYE for Deep Learning (IE7615).
What I learned
- MLOps as engineering: versioned data and models, release gates and monitoring.
- Evaluating models honestly: baselines, confidence intervals and decisions made before the results.
Fidelity
Experience, Fidelity Investments, Boston, Jul – Dec 2025. Data Analyst Co-op.
A new Health & Welfare business with no shared numbers. I unified its scattered systems into one source of truth.
1.3M employees unified; 3M+ enrollments; 12M+ engagements a year.
What I did
- Built the first standardized reporting framework and metrics for the Health & Welfare offering.
- Unified siloed legacy systems into one Snowflake source of truth, with accuracy, completeness and consistency checks.
- Designed Tableau Factbook dashboards across employees, plan sponsors, enrollments and engagement, using correlation, key-driver and trend analysis over 50+ plan sponsors.
- Translated the insights for Sales, Marketing, Health and senior leadership.
What I learned
- One agreed definition of a metric is worth more than ten dashboards.
- Data quality checks are what earn leadership’s trust in the numbers.
- The same insight needs telling differently to Sales, Marketing and leadership.
MOMENTO
Project, Northeastern, with my team, Jan – Apr 2026. Reader compatibility platform, MLOps.
A reading app that matches people by how they think and feel about the same passage. I originated it, designed the ML and built the front end with my team.
2nd of 30 teams, Google-sponsored MLOps Expo; 450 interpretation benchmark.
What I did
- Agent-based compatibility engine on Gemini 2.5 Flash with a Bradley-Terry ranking model.
- Three-metric release gate backed by an offline evaluation harness.
- PII, injection, profanity and spam checks that trigger retraining when metrics regress.
What I learned
- Evaluate before you ship: the offline harness and release gates caught regressions early.
- Leading a team product from an idea to a working demo.
THE EYE
Project, Deep Learning, Northeastern, Sep – Oct 2026. Explainable object detection.
Finds 73 everyday objects, and explains every mistake it makes: wrong ID, missed or extra.
6,942 photos labeled; 98.9% box accuracy; 5 models compared.
What I did
- Human-in-the-loop labeling: OWLv2 prompts plus a self-trained YOLO, combined by majority vote, with disagreements sent to review.
- Model choices made by pre-registered, SHA-256-sealed rules; every result reported with bootstrap confidence intervals.
- Deployed on Oracle Cloud with HTTPS, auto-restart and per-user rate limiting.
What I learned
- I turned down a five-model ensemble that got 6.9 points more images fully correct, because it lost recall on cut-out scenes and added false boxes.
- Pre-register decisions so the results can’t steer them.
- Report uncertainty with intervals, not single numbers.
