The Lab: Not everything starts as a product.
The Lab is where I test hypotheses, benchmark new AI models, experiment with on-device inference, and document raw discoveries before packaging them into software.
Autonomous Niche Opportunity Scraper & Viability Classifier
Can an autonomous AI agent scrape unranked marketplace niches, analyze 1-star review clusters, and validate search velocity to propose viable digital micro-products in under 5 minutes without human intervention?
Constructed a 4-node LangGraph agent with Playwright headless scraper, Cohere embedding clustering, and Gemini 2.0 Flash unit economics evaluator. Ran across 40 sub-categories in utility software and mobile productivity.
Agent successfully isolated 12 high-intent unmet software features with under 3 competitive solutions. Identified recurring complaints around pet health coordination and niche technical exam preparation.
Pure keyword volume is a lagging indicator. The highest signal lies in negative review sentiment clusters with phrases like 'I wish this app also had...' or 'Impossible to do X with my spouse'.
Sub-200ms Latency Audio Pipeline for Real-Time Language Coaching
Can we achieve natural, conversational turn-taking (< 200ms latency) in Italian language voice coaching using WebSockets, streaming Whisper STT, and small quantized local TTS models?
Benchmarked direct client WebSocket streaming vs Server-Sent Events with edge proxying. Tested OpenAI Realtime API vs modular pipeline (Whisper Edge -> Groq LLaMA-3-8B -> Kokoro TTS).
The modular edge pipeline achieved 230ms average glass-to-glass latency, while OpenAI Realtime API hit 180ms. Kokoro TTS provided natural Italian cadence with minimal server footprint.
Voice coaching doesn't just need fast latency; it requires conversational backchanneling ('um', 'certo', 'esatto') to prevent students from feeling awkward while thinking in a foreign language.
Dynamic Prompt Reranking vs. Fine-Tuning for Bureaucratic Legal Text
Does dynamic few-shot retrieval (RAG with semantic similarity) outperform an instruction fine-tuned 8B model when explaining tricky Italian highway legal rules?
Created a benchmark test suite of 400 'trucchi e tranelli' (trick questions) from the Italian Ministry of Transport. Evaluated a fine-tuned Mistral-7B against GPT-4o with dynamic vector retrieval of Codice della Strada articles.
Dynamic few-shot RAG achieved 96.2% factual legal accuracy versus 84.1% for the fine-tuned model, while eliminating fine-tuning maintenance when legal amendments are passed.
For regulatory domains where laws change and exact wording matters, RAG + strong reasoning models consistently defeats fine-tuning, which suffers from subtle parametric hallucinations.
On-Device SLM (Small Language Model) for Offline Health Symptom Logging
Can a 1.5B parameter quantized model (like Qwen2.5-1.5B or Gemma 2 2B) run on mid-range Android hardware to parse voice symptoms into structured health JSON offline?
Tested ONNX Runtime and MediaPipe GenAI on Snapdragon 7-series and Tensor G2 chips with 4-bit INT4 quantization. Measured battery drain and memory footprint.
Gemma 2 2B INT4 ran at 14 tokens/sec with ~650MB RAM footprint. Capable of parsing free-form symptom descriptions into structured JSON tags without internet connectivity.
Mobile hardware is now fast enough for zero-cloud, privacy-first AI apps. Pet health and personal notes can stay 100% on the device.
Adversarial Multi-Agent Debate for Hallucination Suppression
Does a two-agent adversarial debate (Generator vs Skeptic) significantly reduce hallucination rates in technical code generation compared to single-pass CoT?
Set up 2 agents: Agent A produces code and explanations; Agent B acts as a senior reviewer with strict linter and test case execution tools. Evaluated on 100 tricky edge-case programming briefs.
Syntax and logic regressions dropped by 73%. Agent B caught 91% of edge cases (null checks, race conditions) before output delivery.
Agents need separated roles with asymmetric incentives. Giving the reviewer agent a 'bounty' for finding bugs produces far better results than asking a single agent to 'review its own work'.