Susmit Vengurlekar
|
Specialized in real-time data pipelines (Kafka, Flink), graph databases (Neo4j), and full-stack development. I design and build end-to-end solutions from backend data infrastructure to intuitive user interfaces.
KEY SKILLS
SPEAKER
Featured Work
Real-Time Click-Through Rate Analysis
The Challenge
High latency in processing user impressions and clicks for ad-tech analytics.
The Solution
Implemented a Flink streaming job with Kafka as the message bus to process data in real-time.
Impact
Reduces data processing latency from minutes to milliseconds, enabling real-time decision making.
Kong API Gateway with Observability
The Challenge
Lack of visibility into microservices traffic and performance bottlenecks.
The Solution
Deployed Kong API Gateway with OpenTelemetry and OpenObserve for end-to-end tracing and metrics.
Impact
Improves system reliability and reduces mean time to resolution (MTTR) for API issues.
QnA on Knowledge Graph
The Challenge
Difficulty in retrieving accurate answers from unstructured data using traditional search.
The Solution
Built a RAG system using Neo4j as a knowledge graph to provide context-aware answers.
Impact
Enhances answer accuracy and relevance by leveraging graph relationships.
Bonus Payout Curve Generator
The Challenge
Designing a fair bonus payout curve that hits target budget utilization while keeping engagement in a desired band - a high-dimensional, non-convex tuning problem with six interdependent parameters (start/mid/max points and bonuses).
The Solution
Built a parametric payout generator combining an exponential fit (start→mid) with a saturation-growth fit (mid→max), then wrapped it in a DEAP genetic algorithm that drives budget utilization toward 100% (minimizing |100 − utilization|) under engagement-band penalties. Added a sensitivity analysis pass for ±10% sales scenarios.
Impact
Optimizer lifted budget utilization from 85.5% to 93.4% and engagement rate to a clean 90.0% while keeping the curve monotonic and within stakeholder bounds - turning a manual, judgment-heavy exercise into a reproducible, data-driven calibration.
Scalable Multi-Engine Data Profiler
The Challenge
To generate highly accurate SQL, AI agents need deep context about a database (schemas, value distributions, cardinality). However, extracting and unifying this metadata across completely different data warehouses (Snowflake, Databricks, local DBs) is complex, prone to timeouts on massive schemas, and results in mismatched data types.
The Solution
Architected a resilient, concurrent Python profiling engine using the Adapter Design Pattern. It dynamically dispatches queries to Snowflake, Databricks SQL, DuckDB, or SQLite. It pushes heavy statistical aggregations down to the warehouse, harmonizes engine-specific types into a unified taxonomy, and features a robust checkpoint/resume system to handle network interruptions.
Impact
Produces deterministic, highly structured JSON artifacts containing precise database intelligence (histograms, exact/approximate distinct counts, null ratios). This standardized context dramatically improves LLM text-to-SQL accuracy while optimizing warehouse compute costs.
Work Experience
Data Scientist & Solution Architect at SkillRev
- Integrating AIDEN AI capabilities into client solutions to enhance business insights.
Founding Team Member at AIDAX
- Responsible for the strategic direction and development of AIDAX, focusing on innovative AI solutions.
Senior Data Scientist and Software Architect at Xcellen PTE Ltd
- Developed Gen AI use cases, including identifying logic breaches in survey data and automating insights generation for PowerPoint slide decks.
- Automated the creation of Diagnostic Framework PowerPoint presentations with thinkcell charts, text, and tables based on tabular data.
- Prepared and documented user, logic, and data flow diagrams; developed a Bonus Payout Curve Generator; acted as technical program manager for the in-house product XBoost.
Team Lead and Data Scientist at Zeza Technologies
- Led a team of 4 developers and 3 interns, coordinating with QA Head and Project Manager to delegate tasks effectively.
- Assumed additional roles such as Frontend Developer and DevOps as needed.
- Planned features in advance, breaking them down into well-defined and documented tasks for estimation and progress measurement.
Data Scientist and Backend Developer at Zeza Technologies
- Developed a Data Engineering and ML-as-a-Service platform with auto ML, auto feature engineering, EDA, and explainable AI capabilities.
- Reduced costs and improved performance by ~60% by rearchitecting systems using SQS, Circle CI, on-demand EC2s, Lambda functions, auto timeout systems, and static site hosting.
Backend Developer and Database Engineer for Flyer Lively: Interests & Hobbies
- Developed backend services and managed databases using PostgreSQL, Node.js, Python, and the Serverless Framework.
Data Scientist & Solution Architect at SkillRev
- Integrating AIDEN AI capabilities into client solutions to enhance business insights.
Founding Team Member at AIDAX
- Responsible for the strategic direction and development of AIDAX, focusing on innovative AI solutions.
Senior Data Scientist and Software Architect at Xcellen PTE Ltd
- Developed Gen AI use cases, including identifying logic breaches in survey data and automating insights generation for PowerPoint slide decks.
- Automated the creation of Diagnostic Framework PowerPoint presentations with thinkcell charts, text, and tables based on tabular data.
- Prepared and documented user, logic, and data flow diagrams; developed a Bonus Payout Curve Generator; acted as technical program manager for the in-house product XBoost.
Team Lead and Data Scientist at Zeza Technologies
- Led a team of 4 developers and 3 interns, coordinating with QA Head and Project Manager to delegate tasks effectively.
- Assumed additional roles such as Frontend Developer and DevOps as needed.
- Planned features in advance, breaking them down into well-defined and documented tasks for estimation and progress measurement.
Data Scientist and Backend Developer at Zeza Technologies
- Developed a Data Engineering and ML-as-a-Service platform with auto ML, auto feature engineering, EDA, and explainable AI capabilities.
- Reduced costs and improved performance by ~60% by rearchitecting systems using SQS, Circle CI, on-demand EC2s, Lambda functions, auto timeout systems, and static site hosting.
Backend Developer and Database Engineer for Flyer Lively: Interests & Hobbies
- Developed backend services and managed databases using PostgreSQL, Node.js, Python, and the Serverless Framework.
Education
B.Sc. in Information Technology from DG Ruparel College, Mumbai University (2018 - 2021) - CGPA: 9.73
HSC Commerce from D.G. Ruparel College of Arts, Science and Commerce - 86.46%
FYJC (11th Std.) - 82.15%
SSC - 91.4%
Tech Blogs
Technical insights and deep dives
Live Projects
Hover over card to see demo
GitHub Repositories
Open source contributions and experiments
Tech Stack
Expert tools and usecase matchmaker
- Python
Dart- TypeScript
sklearn
- Data Engineering
- Machine Learning
- PostgreSQL
Serverless Framework
GitHub- Docker
Professional Courses and Badges
Verified credentials and certifications
Introduction to Stream Processing and Apache Flink®
Neo4j Graph Data Science Fundamentals
MongoDB Advanced Schema Design Patterns and Anti-patterns Skill Badge
MongoDB Aggregation Fundamentals
Building AI Agents with MongoDB
Building RAG Apps Using MongoDB
Building AI-Powered Search with MongoDB Vector Search
Introduction to Apache Flink® SQL
MongoDB Query Optimization Techniques
Neo4j - Introduction to Vector Indexes and Unstructured Data
Introduction to Stream Processing and Apache Flink®
Neo4j Graph Data Science Fundamentals
MongoDB Advanced Schema Design Patterns and Anti-patterns Skill Badge
MongoDB Aggregation Fundamentals
Building AI Agents with MongoDB
Building RAG Apps Using MongoDB
Building AI-Powered Search with MongoDB Vector Search
Introduction to Apache Flink® SQL
MongoDB Query Optimization Techniques
Neo4j - Introduction to Vector Indexes and Unstructured Data
What People Say
“I entered in knowing absolutely nothing but you never treated me like a novice, you shared your intellect with me just as any other developer, and I won't ever forget that. You also brought the fun memories in Zeza, and I have never seen anyone else with this much knowledge and experience be so down to earth and funny. All in all, you have inspired me, and I won't ever forget what I learned through you.”
Let's Connect
I'm always enthusiastic to get exposed to different problems and challenges.
Recommended Books
Books that shaped my thinking