Revenue Forecasting · Apr 2022 – Present
Revenue Forecasting Platform
Predict quarter-end revenue every day from evolving pipeline and booking data, where a forecast that looks good offline can still fail badly at quarter end.
AI Engineer · Bengaluru · Open to opportunities
Akash Sharma
AI Engineer
Building intelligent systems that survive contact with real-world data.
I work across applied machine learning, forecasting, data engineering, model validation, and production systems, taking problems from messy data and uncertain assumptions to systems people can actually rely on.
Selected signals
- 4+yearsBuilding ML systems from prototype to production.
- 25+systemsDesigned and shipped across forecasting, attribution, propensity, and data engineering.
- 8+model familiesCombined inside one pipeline projection engine.
- 6h → 1hmodeling runtimeCut by parallelizing model training with joblib.
- 5–15%model MAPEAcross customers, in daily end-of-quarter forecasting.
How I think
From messy data to a system that holds up.
- 01
Frame the problem
Understand the business problem, data-generating process, constraints, and what “good” actually means.
- 02
Find the signal
Explore distributions, semantics, leakage, temporal behavior, missingness, and feature quality before choosing a model.
- 03
Validate the idea
Use leakage-safe validation, appropriate metrics, baselines, and experiments to check whether the model is learning something useful.
- 04
Engineer the system
Turn the validated idea into a configurable, explainable, scalable production system.
- 05
Observe reality
Monitor predictions, edge cases, failures, drift, customer behavior, and unexpected system interactions.
- 06
Improve
Iterate based on evidence rather than assumptions.
Selected work
Systems I've built
Marketing Science · 2022 – Present
Marketing Mix Modeling Platform
Estimate how each marketing channel contributes to pipeline, so budget decisions rest on evidence rather than habit.
Marketing Science · 2022 – 2024
Multi-Touch Attribution System
Credit each channel for its real influence on conversion across a customer journey, not just the last touch.
Pipeline Intelligence · Apr 2022 – Dec 2024
Sales Pipeline Forecasting
Aggregate record-level scores from many model families into daily multi-quarter projections that revenue teams can plan against.
Where I've built
Production ML systems I've owned, end to end.
4+ years building, shipping, and scaling data science and AI systems in B2B SaaS, from initial prototypes to production platforms.
Revsure AI
AI Engineer
CurrentRevsure AI
Dec 2024 – Present · 1.5+ years
- Holds model MAPE in the 5–15% range through forecast category feature engineering, with daily EOQ forecast accuracy of ~85–99%
- Deployed time-decay and average-index forecast adjustment layers to production
- Built dual-algorithm explainability infrastructure (SHAP + coefficient contribution) with BigQuery logging
- Improved configurable multi-model ML framework to support 30+ parameters
- Fixed critical MMX production bugs: MinMax extrapolation, scenario cascade drift, STL edge cases
- Authored comprehensive methodology documentation and customer-facing guidance
AA
Data Scientist
ADA Asia
Apr 2022 – Dec 2024 · ~2 years 9 months
- Built the Revenue Forecasting Platform from QTD heuristic through production XGBoost-based EOQ system
- Engineered the Marketing Mix Modeling Platform from research through production
- Designed and owned the Sales Pipeline Forecasting system integrating 8+ ML model families
- Built Markov Chain Multi-Touch Attribution system with firmographic and campaign metadata enrichment
- Migrated revenue metrics pipeline from pandas to distributed PySpark, resolving critical implementation bugs
- Reduced modeling pipeline from 5–6 hours to 1 hour via joblib parallelization
- Provided guidance and mentorship to interns across the team
Education
IL
Post Graduate Program in Data Analytics
Imarticus Learning
Jan 2022 · Thane, India
TP
Bachelor of Information Technology
T.Z.A.S.P. Pragati College
Nov 2020 · Dombivli, India
Verified Credentials
11 active IBM × Coursera credentials verified via Credly (earned 2021).
IBM Data Science Professional Certificate
IBM × Coursera · Oct 2021
Applied Data Science Capstone
IBM × Coursera · Oct 2021
Machine Learning with Python
IBM × Coursera · Sep 2021
Data Analysis with Python
IBM × Coursera · Jun 2021
Data Visualization with Python
IBM × Coursera · Sep 2021
Databases and SQL for Data Science
IBM × Coursera · Apr 2021
Technical depth
Tools, grouped by what I use them for.
Depth is shown in the projects above. This is the working toolkit behind them.
Modeling
- XGBoost
- scikit-learn
- BayesianRidge
- SHAP
- statsmodels
- SciPy
Forecasting
- EOQ forecasting
- Multi-horizon forecasting
- STL
- SARIMAX
- Temporal validation
Data systems
- Python
- SQL
- BigQuery
- PySpark
- pandas
- NumPy
Production
- Airflow
- GCP
- Dataproc
- Cloud Storage
- APIs
- Monitoring
- Explainability
Domain
- Revenue intelligence
- Pipeline forecasting
- Marketing attribution
- Marketing mix modeling
- GTM analytics
Currently exploring
I've spent the last several years building production ML systems. I'm now extending that systems mindset into modern AI.
- LLM orchestration
- Multi-agent systems
- Advanced RAG
- AI evaluation
- Observability
Personal lab · experiments · builds
Akash Labs
Work is what I shipped. Lab is what I'm exploring: rougher on purpose, and not a company.
Narovva: autonomous news-to-social publishing engine
Discovers news, structures an LLM brief, renders branded carousels deterministically, and publishes to Instagram on a schedule. The LLM handles article understanding and captions; layout, scheduling, and failure handling stay deterministic and testable.
- AI experiments
- Generative AI
- Data experiments
- Visualization
- Tools
- Prototypes