London, United Kingdom

Data scientist

Aashish Biradar

I build machine learning that gets used: NLP, computer vision and time-series forecasting, from the data pipeline to the dashboard people act on. Born in Gulbarga, India, grew up in Sohar, Oman, studied and worked in India, and now in London for an MSc in Finance Analytics at King's College London.

Available now in the UK on a Graduate visa (no sponsorship needed for 2 years).

Download CV LinkedIn
Map: born in Gulbarga, India, grew up in Sohar, Oman, university and work in India, now in London. Solo trips to Japan, Turkey, Canada, Mexico and Peru. 0° JPN TUR CAN MEX PER IND OMN GBR
Where I've livedSolo trips
  1. INDBorn in Gulbarga
  2. OMNGrew up in Sohar
  3. INDDegree and first jobs2018 - 2025
  4. GBRMaster's in London2025 - now

Work

Data Scientist · 2024 - 2025 · Bengaluru

WiseWork

Machine learning and data tooling for an engineering analytics platform.

A support bot that understands intent

accuracy of the NLP intent model, in production
Query handling time
down 25%

Customer satisfaction up 15%.

Error logs, grouped by type

Unsupervised clustering (scikit-learn) on production logs from Supabase, 30+ incidents a month.

Time to resolve, before10h
After7h

Exam sheets, read by a CNN

  1. Scan
  2. Crop the roll-number box
  3. CNN reads the digits
  4. Marks database

Replaced manual paper correction.

Safer releases

Regressions after release
down 40%

A/B validation and monitoring on every model update. Backend services, REST APIs and CI/CD in Python and C#/.NET.

Graduate Engineering Trainee · 2022 - 2023 · Bengaluru

TVS Motor Company

Data science and forecasting for EV manufacturing.

EV charge-cycle forecasts, 25% more accurate

+25%forecast accuracy
TrainTest, always later in time

Expanding-window, out-of-sample validation, so the model is never tested on data it has seen. Senior leadership adopted the dashboards for production and capacity planning.

Payroll analytics pipelines

  1. SAP
  2. Azure Data Factory
  3. Data Lake
  4. Power BI
40%faster processing
600+hours saved a quarter
1,000+employees, more accurate pay data

Raw-material prices, tracked automatically

Manual procurement work
halved

Python, feeding weekly sourcing decisions.

MSc dissertation · King's College London · 2025 - now

Does news sentiment help forecast UK stock returns?

Forecasting the UK equity premium with machine learning and FinBERT.

  1. FTSE All-Share2000 to 2025, 300 months
  2. ~198,000news articles
  3. FinBERTsentiment, beat a classical dictionary
  4. 3 modelselastic net, random forest, gradient boosting
  5. Testedout-of-sample R², Clark-West, backtest with trading costs

The random forest with sentiment beat buy-and-hold on Sharpe ratio, with a smaller drawdown, while holding less equity.

Side project · Oct 2026 - now

Double O Broke

My personal AI agent on Android, built with the help of an AI coding agent.

Money

Bank alerts become categorised spending, tracked against my goals.

Email

Sorts email by what it says (Claude, Gemini), and only pings me when it matters.

Travel

A day of noisy GPS becomes tube and bus trips, priced with TfL's open data.

Screen time

Strict daily limits per app: an app is blocked once its time is up, and a limit can only be loosened from the next day.

30 to 45minutes of manual work saved a day
3.5 ha day in total on Instagram and YouTube
39releases in its first week
6 → 2trips detected on a real day, down to the 2 actually taken
See the code on GitHub
Side project · Oct 2026 - now

Just Let Me Pass This

My AI study agent for the Japanese Language Proficiency Test, built with the help of an AI coding agent, for my N4 exam in December.

Plans the day

One Study now button, paced from the exam date and a daily time goal.

Spaced repetition

A right answer comes back later, a wrong one comes back soon.

Graded stories

From N5 to N1, with furigana that fades as each word is learned.

Claude

Explains wrong answers, chats in Japanese and writes new readings.

8,034words
2,211kanji
526grammar points
254graded stories
See the code on GitHub
B.Tech thesis · NIT Rourkela · 2022

Fake news detection

99%best model
LSTMBi-LSTM
vs
Logistic regressionNaive BayesPassive aggressive

A full NLP pipeline (stemming, word embeddings) on a public news dataset.

People

Leading teams

4
Engineering squadWiseWork, backlog and releases
12
Basketball squadVice-captain, NIT Rourkela
90+
MSc cohortStudent representative, King's

The journey

INDBorn

Gulbarga, India

Born in Gulbarga (Kalaburagi), Karnataka. I moved to Oman when I was 10 months old.

OMNSchool

Sohar, Oman

Grew up in Sohar and went to school there up to 12th grade, then came back to India for university.

IND2018 - 2022

B.Tech Computer Science and Engineering, NIT Rourkela

Computer science, a thesis on fake news detection, and vice-captain of the university basketball team.

IND2022 - 2023

TVS Motor Company

Graduate Engineering Trainee, Bengaluru. Data science and forecasting for EV manufacturing.

My first job, and where I started working closely with finance.

IND2024 - 2025

WiseWork

Data Scientist, Bengaluru. Machine learning and data tooling for an engineering analytics platform.

NLP, clustering and computer vision in production, and leading a small engineering squad.

GBR2025 - now

MSc Finance Analytics, King's College London

King's Business School, London.

Why finance: I've always been excited by data and finding what it really says. In my first job I worked closely with finance, and growing up I watched my dad trade as a hobby, so the interest grew. A Finance Analytics course at King's, in London of all places, was the right next step. I loved the course and learnt a lot.

Student representative for the cohort, and the Finance Trading Program certification.

Off the clock

Passport

Travel log

Five solo trips abroad in just over a year, plus one closer to home.

  1. IND

    Delhi and Agra

    The Taj Mahal, my first of the New 7 Wonders.

  2. JPN

    Japan

    My first solo trip, to the country I had wanted to see for 10 years. Got my Open Water certification in Okinawa and went to my first concert.

  3. TUR

    Turkey

    First trip with my camera. A hot air balloon over Cappadocia, and dives in Fethiye.

  4. CAN

    Canada

    Visited a friend from high school, tried venison for the first time, and picked up the DJI camera.

  5. MEX

    Mexico

    Advanced Open Water, diving in Cancún, and Chichén Itzá.

  6. PER

    Peru

    Machu Picchu. Loved the mountains, the culture and the views.

3 of 7 New 7 Wonders of the World

  1. 1Taj MahalIndia · Jul 2025
  2. 2Chichén ItzáMexico · Sep 2026
  3. 3Machu PicchuPeru · Oct 2026
  4. Great WallChina
  5. PetraJordan
  6. ColosseumItaly
  7. Christ the RedeemerBrazil
Water

Scuba diving

Favourite dive site so far: Cancún.

PADIOpen Water DiverJapan · Aug 2025
PADIAdvanced Open Water DiverMexico · Sep 2026
18 m25 m, my deepest30 m
0 m102030
14dives
7 h 35underwater
4countries
Dive log: depth of each of 14 dives Bars hang down from the surface to each dive's maximum depth. Okinawa, August 2025: 8.1, 8.4, 7.8, 14.7 and 18 metres. Fujairah, September 2025: 10 and 13 metres. Fethiye, July 2026: 12 and 10 metres. Cancún, September 2026: 10, 10, 15, 10 and 25 metres, the deepest, on Wreck C-58. 10 m 20 m 30 m 0 m 25 m, Wreck C-58 Okinawa2025 Fujairah2025 Fethiye2026 Cancún2026
All 14 dives
DateSiteDepthTime
1 Aug 2025Okinawa, Japan8.1 m25 min
1 Aug 2025Okinawa, Japan8.4 m33 min
2 Aug 2025Okinawa, Japan7.8 m20 min
2 Aug 2025Okinawa, Japan14.7 m34 min
2 Aug 2025Okinawa, Japan18 m33 min
13 Sep 2025Fujairah, UAE10 m30 min
13 Sep 2025Fujairah, UAE13 m35 min
15 Jul 2026Fethiye, Turkey12 m40 min
15 Jul 2026Fethiye, Turkey10 m30 min
22 Sep 2026MUSA, Isla Mujeres10 m40 min
22 Sep 2026Manchones, Cancún10 m35 min
23 Sep 2026Punta Negra reef15 m40 min
23 Sep 2026Santa Remedia10 m35 min
24 Sep 2026Wreck C-58, Cancún25 m25 min
Eyes

Photography

Mostly street photography. What's in the bag:

Canon EOS R5024.2 MP APS-C mirrorless
Sigma 18-50mm F2.8f/2.8 all the way
Canon RF-S 18-45mmf/4.5-6.3
1525354555 mm
DJI Osmo Action 5 ProWaterproof to 20 m, no case needed
Words

Languages

  • EnglishEN
  • हिन्दीHI · Hindi
  • 日本語JA · Japanese, basic. Taking JLPT N4 in December.
Lens

Through my lens

Shot on the R50.

Fireworks at Niagara Falls, Canada
Fireworks at Niagara Falls, Canada
Evening light on the Thames, London
Evening light on the Thames, London
The London Eye at golden hour
The London Eye at golden hour
Plane trees
Plane trees
The Shard at sunset, London
The Shard at sunset, London
Lamp and a daytime moon
Lamp and a daytime moon
Take-off, Turkey
Take-off, Turkey
Hands

Speedcubing

Runner-up in a competition. The shelf:

18s3×3 best single
~22sin competition
2×2
3×3
4×4
5×5
Mirror
Air

Skydiving

Jumped out of a plane over Dubai.

Court

Basketball

Vice-captain at NIT Rourkela, running a 12-member squad.

Tools

Programming
Python (Pandas, NumPy, scikit-learn, TensorFlow, PyTorch), SQL (MySQL, Supabase), C#
Machine learning
NLP, transformers (FinBERT), gradient boosting, random forests, time-series forecasting, deep learning (CNNs, LSTM), A/B testing
Evaluation
Out-of-sample R-squared, Clark-West test, expanding-window validation, backtesting
Data
Azure Data Factory, Azure Data Lake, ETL pipelines, Power BI, Flask REST APIs
Ways of working
Git, CI/CD, code review, Agile, experiment design

Say hello

Open to new roles and good conversations. The quickest way to reach me is email.

aashish.biradar13@gmail.com Download CV LinkedIn GitHub