Open to roles — Toronto / Remote

Shivam Bhosale turns
data into decisions &
decisions into intelligent systems ↗

Data Analyst by day, AI Engineer by craft. 3+ years owning the pipeline end-to-end — SQL & Python for analytics, Power BI for storytelling, and LLM / deep-learning systems for the hard problems. Currently shipping ML in production at MJR Capital and building agentic AI tools with Claude + MCP on the side.

Based in
Toronto, ON
Stack
Python · SQL · Power BI · Claude · AWS
Specialty
Analytics · ML in production · Agentic AI
Available
Full-time · Hybrid / Remote
Now / current role Data Analyst → shipping ML & agents +2.4% ▲
01 / Selected work

Analytics & AI that shipped.

(01)
LLM · Agentic AI
Bloomberg Terminal Replica
Claude APIPythonYahoo FinanceECB FXPrompt engineering
AI-powered market intelligence dashboard. Claude reads equities, FX, and macro signals from free APIs and writes a daily institutional-grade brief — for retail users, at zero cost.
Why it exists: Bloomberg is $2k/mo. This gets 80% of the signal with an LLM and free data.
  • Structured-output prompting for reliable parsing
  • Tool use: Claude reasons over live feeds
  • Democratizes institutional analytics
↗
(02)
AI Developer Tool · Gumroad
BayStreetBot
PythonMulti-factor modelsPandasProductized
Productized AI developer tool on Gumroad — a quantitative TSX screener applying multi-factor filters. End-to-end exercise in taking an AI/analytics workflow from notebook to paying customer.
What I learned: productizing AI is mostly packaging, docs, and distribution — not the model.
  • Ships as a single .py — zero setup
  • Multi-factor quant logic: P/E, volume, 52-wk, momentum
  • Live on Gumroad — real paying users
↗
(03)
Agentic Pipeline · LLM
Tech & Science News Aggregator
Gemini FlashPythonRSS / feedparserNewsData.ioCategorization
Agentic news pipeline: ingestion → dedupe → LLM summarization → categorization across AI/ML, Space, Medicine, Climate, and Defence. Gemini Flash free tier runs it at zero cost.
Hard part: deduping across feeds without losing signal, keeping summarization cheap.
  • Title + URL hashing with Levenshtein fallback
  • Summary cache — same story costs nothing twice
  • LLM-based category assignment, not keyword rules
↗
(04)
Deep Learning · Computer Vision
CNN Plant Disease Classifier
TensorFlowKerasDjango35K imagesFederal research
Federally partnered research with Agriculture & Agri-Food Canada. Trained CNNs on 35,000+ annotated images; deployed real-time inference via Django with role-based access for government stakeholders.
Outcome: 95% classification accuracy — a 15% gain via hyperparameter tuning and preprocessing.
  • Rigorous AI lifecycle: EDA → training → evaluation → deploy
  • Explainability reports for non-ML regulatory audiences
  • Production Django inference endpoint with RBAC
↗
02 / Experience

Where I've worked.

2025 — Present
Data Analyst / AI Engineer at MJR Capital Services
Current role: end-to-end financial analytics — Python/SQL ETL, Power BI dashboards with DAX — layered with production ML (classification, regression, anomaly detection) on AWS. Model outputs feed back into executive dashboards via DAX for real-time risk monitoring.
40%
Reporting time cut
Prod
ML models live
2023 — 2024
AI Engineer at U. of Windsor × AAFC
Graduate research with Agriculture & Agri-Food Canada. Trained CNNs on 35K+ images; deployed real-time inference as a Django app with role-based access; authored explainability reports for federal stakeholders.
95%
Model accuracy
35K
Images trained
03 / Capabilities

The stack I reach for.

Analytics & SQL
a.01
  • Python — Pandas, NumPy
  • SQL (advanced)
  • EDA · feature engineering
  • PySpark · R
BI & Reporting
a.02
  • Power BI · DAX
  • Executive dashboards
  • Power Automate
  • Tableau
LLMs & AI APIs
a.03
  • Anthropic Claude API
  • OpenAI · Gemini Flash
  • Prompt engineering · tool use
  • Structured outputs
Agentic AI
a.04
  • LangGraph · CrewAI
  • MCP · A2A protocols
  • RAG pipelines
  • Google ADK
ML & Deep Learning
a.05
  • Scikit-learn — classification, regression
  • TensorFlow · PyTorch · Keras
  • CNNs · anomaly detection
  • Hyperparameter tuning
AI App Development
a.06
  • Django · FastAPI · Flask
  • Real-time inference endpoints
  • REST APIs
  • Production deployments
Cloud & MLOps
a.07
  • AWS — S3, EC2
  • Docker · CI/CD
  • Azure · GCP
  • Model lifecycle · pipelines
Currently exploring
a.08
  • Multi-agent orchestration
  • Long-context RAG
  • Eval harnesses · LLM-as-judge
  • dbt · modern data stack
04 / Contact

From dashboards to
intelligent systems.

Email
shivambhosale909
@gmail.com
Fastest reply — always read.
LinkedIn
/in/shivambhosale
Open to DMs from recruiters.
GitHub
@ShivamBhosale
Code · side projects · PRs.
Location
Toronto, ON
Hybrid or remote.

Tweaks