I believe engineers are not merely computer scientists, philosophers, physicists, or analysts. We are humans above everything and what makes us human is our ability to create things that improve the lives of others. We are a mosaic of our knowledge, applying what we know like painting with colors we have on our palette, to blend into one masterpiece. A solution.
Shreyash S. Bhatkar
Computer Engineer • Software Developer • AI/ML • Cloud-Native
I build reliable systems and applied ML solutions. Below are selected projects and a quick peek at my experience.
Work Experience
A quick glance at my professional background.
Oracle India — Associate Consultant
Jul 2023 – Present
Java Developer — Batelco (Jan 2025 – Present)
- Architected 30+ Java/Kotlin microservices and REST endpoints on OCI.
- Built LangChain + Cohere docs agent to reduce API onboarding time by ~40%.
- Optimized REST/gRPC with 5G core (NRF, PCF) & Wi-Fi control; throughput +37%, latency −12% (Prometheus).
Full-Stack — PPC Retail (Jul 2023 – Dec 2024)
- 6+ reusable BRM opcodes; ~35% increase in billable profit/allocated resource.
- 13+ billable solutions using OJET & Node.js; module load time −23%.
- SSR/serverless with 30+ reusable APIs; backend load −17% (Prometheus).
Nibha Tech — AI Research Intern during Bachelors
Dec 2022 – May 2023
- 30k+ feature vectors over 25 years of data; SARIMA + MACD + Bollinger + SHAP.
- GANs for synthetic data; Backtrader for backtests; MLflow benchmarks — 96.2% normal / 76.03% stress.
- Integrated 15GB+ pipelines; real-time prediction APIs.
Featured Projects
Selected work with outcomes & tech details.
- Piezo-triboelectric energy harvesting ~10–50 mW/stride (bench-verified).
- ARM-M0 + BLE board (32 kB) powering non-lethal actuator from EDLC + Li-Po hybrid stack.
- Captured 40k+ footfall samples (FlexiForce + Data Logger) for training.
- On-device RL (DQN/PPO) classifies routine vs high-risk; 93% F1; triggers only if risk > 0.8.
- Indian Provisional Patent App. No. 2025/XXXXXX — filed Mar 2025 (full spec in preparation).
Real-time Cloud Performance Analytics & Optimization App
- Python Flask (CORS) ingesting 8+ GB real-time logs; Spark MLlib predictive analytics.
- +22% resource efficiency; downtime −15%; OJET + Node.js dashboard with −33% load time.
- Deployed on Oracle Cloud (OKE) for scalability and performance.
More Projects
A few more experiments & apps.
Performance Evaluation and Prediction Tool (PEP)
Evaluates product performance, predicts challenges, and highlights improvement areas with scenario forecasts.
Hybrid Emergency Audio Analyzer
Emergency audio interaction analysis and categorization.
CarBMI
DL study: carbohydrate intake vs. workout and physical performance.
R3ND3R
3D model rendering desktop app with Blender-authored supershapes (Python core).
eSKIMo
Text skimmer & labeling tool (TensorFlow + React).
CryptOpus
Web3 NFT/Crypto market interactions with a ledger of transactions.
BuRD
Bus Route Detector & Classifier for visually impaired users.
Black Box
UV sterilizer box with stepper-mounted absorbent; BT control via Arduino.
FireCart
Full shopping app: cart, checkout, and typed forms; transactions ready.
EmbedroW
t-SNE / UMAP / PCA projector for custom datasets (see demo in repo).
Spacebar (Coming Soon)
Self-learning shuttle that learns to land autonomously.
Motion DA
Browser-based motion capture driving a live animated character.
EdibleInsight
Food classification ML aiming to outperform DEEPFOOD (2016).
RollBall
Endless-runner game built with Pygame.
Publication & Patents
Find More
Suggestions, feedback, or just say hi.